BigAudioBench / Dynamic-SUPERB

Sources

1. Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks — published result table ↗

   Huang, Chien-yu, Chen, Wei-Chih, Yang, Shu-wen, Liu, Andy T., Li, Chen-An, Lin, Yu-Xiang et al. · 2024

   All 80 authors

   Huang, Chien-yu, Chen, Wei-Chih, Yang, Shu-wen, Liu, Andy T., Li, Chen-An, Lin, Yu-Xiang, Tseng, Wei-Cheng, Diwan, Anuj, Shih, Yi-Jen, Shi, Jiatong, Chen, William, Yang, Chih-Kai, Ren, Wenze, Chen, Xuanjun, Hsiao, Chi-Yuan, Peng, Puyuan, Wang, Shih-Heng, Kuan, Chun-Yi, Lu, Ke-Han, Chang, Kai-Wei, Ritter-Gutierrez, Fabian, Huang, Kuan-Po, Arora, Siddhant, Lin, You-Kuan, Chuang, Ming To, Yeo, Eunjung, Chang, Kalvin, Chien, Chung-Ming, Choi, Kwanghee, Wang, Jun-You, Hsieh, Cheng-Hsiu, Lin, Yi-Cheng, Yu, Chee-En, Chiu, I-Hsiang, Guimarães, Heitor R., Han, Jionghao, Lin, Tzu-Quan, Lin, Tzu-Yuan, Chang, Homu, Chang, Ting-Wu, Chen, Chun Wei, Chen, Shou-Jen, Chen, Yu-Hua, Cheng, Hsi-Chun, Dhawan, Kunal, Fang, Jia-Lin, Fang, Shi-Xin, Chiang, Kuan-Yu Fang, Fu, Chi An, Hsiao, Hsien-Fu, Hsu, Ching Yu, Huang, Shao-Syuan, Wei, Lee Chen, Lin, Hsi-Che, Lin, Hsuan-Hao, Lin, Hsuan-Ting, Lin, Jian-Ren, Liu, Ting-Chun, Lu, Li-Chun, Pai, Tsung-Min, Pasad, Ankita, Kuan, Shih-Yun Shan, Shon, Suwon, Tang, Yuxun, Tsai, Yun-Shao, Wei, Jui-Chiang, Wei, Tzu-Chieh, Wu, Chengxi, Wu, Dien-Ruei, Yang, Chao-Han Huck, Yang, Chieh-Chi, Yip, Jia Qi, Yuan, Shao-Xiang, Noroozi, Vahid, Chen, Zhehuai, Wu, Haibin, Livescu, Karen, Harwath, David, Watanabe, Shinji, Lee, Hung-yi

   Dates & evaluation details

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Chord Classification · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · MARBLE Key Detection (GiantSteps - Key) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · HEAR Music Transcription (MAESTRO) · NAR · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · HEAR Music Transcription (MAESTRO) · TER · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · HEAR Percussion Instruments Tonic Classification (Mridangam Tonic) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Instrument Pitch Classification (Nsynth) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Pitch Extraction By Lyrics (CSD) · NAR · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Pitch Extraction By Lyrics (CSD) · TER · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Emotion Classificaiton In Songs (EMOTIFY) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · MARBLE Emotion Detection (MTG) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · HEAR Music Genre Classification (ISMIR04) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · MARBLE Genre Classification (MTG) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · MARBLE Music Tagging (MagnaTagATune) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · MARBLE Music Tagging (MTG) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Music Genre Classification (FMA) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · HEAR Percussion Instruments Classification (Beijing Opera Percussion) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · HEAR Percussion Instruments Stroke Classification (Mridangam Stroke) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Instrument Classification (Nsynth) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Instrument Combination Recognition (OpenMIC-2018) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Instrument Source Classification (Nsynth) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · MARBLE Instrument Classification (MTG) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Singing Automatic MOS Prediction (SingMOS) · NAR · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Singing Automatic MOS Prediction (SingMOS) · MSE · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Singing Automatic MOS Prediction (SingMOS) · KTAU · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Singing Automatic MOS Prediction (SingMOS) · LCC · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Singing Automatic MOS Prediction (SingMOS) · SRCC · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · MARBLE Beat Tracking (ASAP) · NAR · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Music Beat Tracking (ASAP) · NAR · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Music Beat Tracking (ASAP) · Miss Time · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Audio Editing Identification (People’s Speech) · NAR · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Audio Editing Identification (People’s Speech) · ACC · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Scene Fake Detection (ASPIRE) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Audio Deep Fake Detection (LJSpeech, WaveFake, MUSDB18HQ) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Singing Voice Deepfake Detection (CtrSVDD, ACEKiSing, M4Singer) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Audio Duration Prediction (NTUML2021) · NAR · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Audio Duration Prediction (NTUML2021) · MSE · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · HEAR Music Speech Classification (MAESTRO, Librispeech) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Sound Effect Detection (RemFx) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Speech Detection (LibriSpeech-TestClean) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Speech Detection (LibriSpeech-TestOther) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Speech Detection (LJSpeech) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Children Song Transcript Verification (CSD) · WER · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Lyric Translation (SingSet) · Sacre Bleu · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Song Lyric Recognition (SingSet) · MER · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · MARBLE Vocal Technique Detection (VocalSet) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Audio Segment Retrieval (Clotho) · NAR · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Audio Segment Retrieval (Clotho) · IoU · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · HEAR Sound Event Detection (DCASE2016 Task2) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Multi-channel Sound Event Understanding (STARSS23) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Animal Classification (WaveSource-Test) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Bird Sound Detection (Warblrb10k) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Cat Emotion Classification (Cat Sound Classification Dataset V2) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Cornell Birdcall Identification · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Environmental Sound Classification (ESC50-Animals) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · HEAR Beehive States Classification (Beehive States) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Domestic Environment Sound Event Detection (DESED-PublicEval) · NAR · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Domestic Environment Sound EventDetection (DESED-PublicEval) · Event-based F1 · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Emergency Traffic Detection (Large-Scale-Audio-dataset) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Environmental Sound Classification (ESC50-Exterior And Urban Noises) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Environmental Sound Classification (ESC50-Interior And Domestic Sounds) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Environmental Sound Classification (ESC50-Natural Soundscapes And Water Sounds) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Environmental Sound Classification (UrbanSound8K-Urban Noises) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Environment Recognition (ESC50) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · HEAR Environmental Sound Classification (ESC50) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · HEAR Vocal Imitation Classification (Vocal Imitations) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Audio Spatial Distance Prediction (Spatial LibriSpeech) · NAR · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Audio Spatial Distance Prediction (Spatial LibriSpeech) · MEDAE · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · How Far Are You (3DSpeaker) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Sound Position Prediction · NAR · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Sound Position Prediction · Angle Diff · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Dialogue Emotion Classification (Daily Talk) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Emoji Grounded Speech Emotion Recognition (RAVDESS) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Emotion Change Detection (Ravdess) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Emotion Recognition (MELD) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · HEAR Emotion Recognition (CREMAD) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB Emotion Recognition (RAVDESS) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Covid19 Cough Audio Classification (CoughVid) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Environmental Sound Classification (ESC50-Human And Non Speech Sounds) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Human Non Speech Sound Recognition (Nonspeech7k, CommonVoice) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Human Screaming Detection (Environmentdb) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Vocal Sound Recognition (VocalSound) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Phoneme Segment Counting (Librispeech-words) · NAR · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Phoneme Segment Counting (Librispeech-words) · ACC · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Phoneme Segment Counting (Librispeech-words) · Abs Diff · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Phone Segment Counting (VoxAngeles) · NAR · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Phone Segment Counting (VoxAngeles) · ACC · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Phone Segment Counting (VoxAngeles) · Abs Diff · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB Phoneme Recognition (LibriSpeech-TestClean) · PER · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Phoneme Recognition (LibriSpeech-TestOther) · PER · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Phonological Feature Classification (VoxAngeles-Consonant Place Of Articulation) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Phonological Feature Classification (VoxAngeles-Manner Of Articulation) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Phonological Feature Classification (VoxAngeles-Phone) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Phonological Feature Classification (VoxAngeles-Vowel Frontness) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Phonological Feature Classification (VoxAngeles-Vowel Height) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Phonological Feature Classification (VoxAngeles-Vowel Roundedness) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Heteronym Differentiation (HeteronymEn) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · L2 English Accuracy (speechocean762-Ranking) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · L2 English Accuracy (speechocean762-scoring) · NAR · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · L2 English Accuracy (speechocean762-scoring) · PCC · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Multi-lingual Pronunciation Similarity (VoxAngeles) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Accent Classification (Accentdb Extended) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Stress Detection (MIRSD) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Third Tone Sandhi Recognition (NCCU Corpus of Spoken Taiwan Mandarin) · NAR · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Third Tone Sandhi Recognition (NCCU Corpus of Spoken Taiwan Mandarin) · IoU · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · L2 English Fluency (speechocean762-Ranking) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · L2 English Fluency (speechocean762-Scoring) · NAR · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · L2 English Fluency (speechocean762-Scoring) · PCC · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · L2 English Prosodic (speechocean762-Ranking) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · L2 English Prosodic (speechocean762-Scoring) · NAR · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · L2 English Prosodic (speechocean762-Scoring) · PCC · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Prosody Naturalness (ProsAudit-Lexical) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Prosody Naturalness (ProsAudit-Protosyntax) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Spoof Detection (ASVspoof2015) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Spoof Detection (ASVspoof2017) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Deep Fake Voice Recognition (DEEP-VOICE) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Enhancement Detection (LibriTTS-TestClean, WHAM) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Fraud Robocall Recognition (CallHome) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Fraud Robocall Recognition (Promo) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Fraud Robocall Recognition (Robocall) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · HEAR Language Identification (VoxLingua107) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Language Identification (VoxForge) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Age Classification (Common Voice) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Gender Recognition by Voice (Common Voice) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · HEAR Speaker Count Identification (LibriCount) · NAR · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · HEAR Speaker Count Identification (LibriCount) · ACC · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Multi-Speaker Detection (VCTK) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Speaker Counting (LibriTTS-TestClean) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Speaker Verification (LibriSpeech-TestClean) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Speaker Verification (LibriSpeech-TestOther) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Speaker Verification (VCTK) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB Speaker Diarization (Libri2Mix-Test) · NAR · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB Speaker Diarization (Libri2Mix-Test) · DER · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB Speaker Verification (SUPERB Hidden Set) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Noise Detection (LJSpeech, MUSAN-Gaussian) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Noise Detection (LJSpeech, MUSAN-Music) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Noise Detection (LJSpeech, MUSAN-Noise) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Noise Detection (LJSpeech, MUSAN-Speech) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Noise Detection (VCTK, MUSAN-Music) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Noise Detection (VCTK, MUSAN-Noise) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Noise Detection (VCTK, MUSAN-Speech) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Noise Detection (VCTK, MUSAN-Gaussian) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Noise SNR Level Prediction (VCTK, MUSAN-Gaussian) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Noise SNR Level Prediction (VCTK, MUSAN-Music) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Noise SNR Level Prediction (VCTK, MUSAN-Noise) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Noise SNR Level Prediction (VCTK, MUSAN-Speech) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Reverberation Detection (LJSpeech, RirsNoises-LargeRoom) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Reverberation Detection (LJSpeech, RirsNoises-MediumRoom) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Reverberation Detection (LJSpeech, RirsNoises-SmallRoom) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Reverberation Detection (VCTK, RirsNoises-LargeRoom) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Reverberation Detection (VCTK, RirsNoises-MediumRoom) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Reverberation Detection (VCTK, RirsNoises-SmallRoom) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · N-Best Correction (Librispeech-TestOther) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · AAVE Speech Recognition (CORAAL) · WER · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Code-switch Speech Recognition (NTUML2021) · MER · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Code-Switching Speech Recognition (ASCEND) · NAR · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Code-Switching Speech Recognition (ASCEND) · ACC · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Multi-Lingual Speech Recognition (MLS-de) · WER · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Multi-Lingual Speech Recognition (MLS-en) · WER · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Multi-Lingual Speech Recognition (MLS-es) · WER · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Multi-Lingual Speech Recognition (MLS-fr) · WER · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Multi-Lingual Speech Recognition (MLS-it) · WER · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Multi-Lingual Speech Recognition (MLS-nl) · WER · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Multi-Lingual Speech Recognition (MLS-pl) · WER · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Multi-Lingual Speech Recognition (MLS-pt) · WER · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · PTBR Speech Recognition (Common Voice) · WER · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB ASR (LibriSpeech-TestClean) · WER · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · ASR (LibriSpeech-TestOther) · WER · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB OOD ASR (Ar) (Common Voice) · WER · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB OOD ASR (Es) (Common Voice) · WER · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB OOD ASR (Spontaneous) (CHIME6-Test) · WER · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB OOD ASR (Zh) (Common Voice) · CER · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Target Speaker ASR (AMI-test) · WER · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Multi-Speaker Detection (LibriSpeech-TestClean) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Speech Command Recognition (AudioMNIST) · NAR · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Speech Command Recognition (AudioMNIST) · ACC · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Speech Text Matching (LibriSpeech-TestClean) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Speech Text Matching (LibriSpeech-TestOther) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Speech Text Matching (LJSpeech) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Spoken Term Detection (LibriSpeech-TestClean) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Spoken Term Detection (LibriSpeech-TestOther) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Spoken Term Detection (LJSpeech) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB Keyword Spotting (Speech Commands V1-Test) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB Query by Example (Quesst14-Eval) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Stuttering Detection (SEP28k) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Voice Disorder Classification (VOICED) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Conversation Matching · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Dialogue Act Classification (SLUE-HVB) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Dialogue Act Classification (Daily Talk) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Dialogue Act Pairing (Daily Talk) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB Intent Classification (SLURP) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB Intent Classification (SLURP-Action) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB Intent Classification (SLURP-Intent) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Named Entity Localization (SLUE-VoxPopuli) · NAR · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Named Entity Localization (SLUE-VoxPopuli) · F1 · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Named Entity Recognition (SLUE-VoxPopuli) · NAR · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Named Entity Recognition (SLUE-VoxPopuli) · IoU · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Sarcasm Detection (Mustard) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Semantic Textual Similarity (SpokenSTS) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Speech Sentiment Analysis (MELD) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Spoken Digit Arithmetic (AudioMNIST) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB Slot Filling (AudioSnips-Test) · NAR · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB Slot Filling (AudioSnips-Test) · Slot Type F1 · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB Slot Filling (AudioSnips-Test) · Slot Value CER · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Code-Switching Semantic Grammar Acceptability Comparison (CSZS-zh-en) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Nonce Word Detection (sWUGGY) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · PoS Estimation (LibriTTS) · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · PoS Estimation with transcription (LibriTTS) · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Sentence Grammar Acceptability (sBLIMP) · LLM-C · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB Speech Translation (CoVoST2-Test) · Sacre Bleu · GPT-4o evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the GPT-4o evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 5; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Chord Classification · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · MARBLE Key Detection (GiantSteps - Key) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · HEAR Music Transcription (MAESTRO) · NAR · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · HEAR Music Transcription (MAESTRO) · TER · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · HEAR Percussion Instruments Tonic Classification (Mridangam Tonic) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Instrument Pitch Classification (Nsynth) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Pitch Extraction By Lyrics (CSD) · NAR · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Pitch Extraction By Lyrics (CSD) · TER · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Emotion Classificaiton In Songs (EMOTIFY) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · MARBLE Emotion Detection (MTG) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · HEAR Music Genre Classification (ISMIR04) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · MARBLE Genre Classification (MTG) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · MARBLE Music Tagging (MagnaTagATune) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · MARBLE Music Tagging (MTG) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Music Genre Classification (FMA) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · HEAR Percussion Instruments Classification (Beijing Opera Percussion) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · HEAR Percussion Instruments Stroke Classification (Mridangam Stroke) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Instrument Classification (Nsynth) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Instrument Combination Recognition (OpenMIC-2018) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Instrument Source Classification (Nsynth) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · MARBLE Instrument Classification (MTG) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Singing Automatic MOS Prediction (SingMOS) · NAR · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Singing Automatic MOS Prediction (SingMOS) · MSE · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Singing Automatic MOS Prediction (SingMOS) · KTAU · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Singing Automatic MOS Prediction (SingMOS) · LCC · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Singing Automatic MOS Prediction (SingMOS) · SRCC · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · MARBLE Beat Tracking (ASAP) · NAR · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Music Beat Tracking (ASAP) · NAR · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Music Beat Tracking (ASAP) · Miss Time · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Audio Editing Identification (People’s Speech) · NAR · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Audio Editing Identification (People’s Speech) · ACC · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Scene Fake Detection (ASPIRE) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Audio Deep Fake Detection (LJSpeech, WaveFake, MUSDB18HQ) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Singing Voice Deepfake Detection (CtrSVDD, ACEKiSing, M4Singer) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Audio Duration Prediction (NTUML2021) · NAR · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Audio Duration Prediction (NTUML2021) · MSE · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · HEAR Music Speech Classification (MAESTRO, Librispeech) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Sound Effect Detection (RemFx) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Speech Detection (LibriSpeech-TestClean) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Speech Detection (LibriSpeech-TestOther) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Speech Detection (LJSpeech) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Children Song Transcript Verification (CSD) · WER · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Lyric Translation (SingSet) · Sacre Bleu · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Song Lyric Recognition (SingSet) · MER · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · MARBLE Vocal Technique Detection (VocalSet) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Audio Segment Retrieval (Clotho) · NAR · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Audio Segment Retrieval (Clotho) · IoU · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · HEAR Sound Event Detection (DCASE2016 Task2) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Multi-channel Sound Event Understanding (STARSS23) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Animal Classification (WaveSource-Test) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Bird Sound Detection (Warblrb10k) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Cat Emotion Classification (Cat Sound Classification Dataset V2) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Cornell Birdcall Identification · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Environmental Sound Classification (ESC50-Animals) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · HEAR Beehive States Classification (Beehive States) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Domestic Environment Sound Event Detection (DESED-PublicEval) · NAR · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Domestic Environment Sound EventDetection (DESED-PublicEval) · Event-based F1 · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Emergency Traffic Detection (Large-Scale-Audio-dataset) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Environmental Sound Classification (ESC50-Exterior And Urban Noises) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Environmental Sound Classification (ESC50-Interior And Domestic Sounds) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Environmental Sound Classification (ESC50-Natural Soundscapes And Water Sounds) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Environmental Sound Classification (UrbanSound8K-Urban Noises) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Environment Recognition (ESC50) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · HEAR Environmental Sound Classification (ESC50) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · HEAR Vocal Imitation Classification (Vocal Imitations) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Audio Spatial Distance Prediction (Spatial LibriSpeech) · NAR · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Audio Spatial Distance Prediction (Spatial LibriSpeech) · MEDAE · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · How Far Are You (3DSpeaker) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Sound Position Prediction · NAR · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Dialogue Emotion Classification (Daily Talk) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Emoji Grounded Speech Emotion Recognition (RAVDESS) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Emotion Change Detection (Ravdess) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Emotion Recognition (MELD) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · HEAR Emotion Recognition (CREMAD) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB Emotion Recognition (RAVDESS) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Covid19 Cough Audio Classification (CoughVid) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Environmental Sound Classification (ESC50-Human And Non Speech Sounds) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Human Non Speech Sound Recognition (Nonspeech7k, CommonVoice) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Human Screaming Detection (Environmentdb) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Vocal Sound Recognition (VocalSound) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Phoneme Segment Counting (Librispeech-words) · NAR · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Phoneme Segment Counting (Librispeech-words) · ACC · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Phoneme Segment Counting (Librispeech-words) · Abs Diff · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Phone Segment Counting (VoxAngeles) · NAR · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Phone Segment Counting (VoxAngeles) · ACC · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Phone Segment Counting (VoxAngeles) · Abs Diff · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB Phoneme Recognition (LibriSpeech-TestClean) · PER · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Phoneme Recognition (LibriSpeech-TestOther) · PER · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Phonological Feature Classification (VoxAngeles-Consonant Place Of Articulation) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Phonological Feature Classification (VoxAngeles-Manner Of Articulation) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Phonological Feature Classification (VoxAngeles-Phone) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Phonological Feature Classification (VoxAngeles-Vowel Frontness) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Phonological Feature Classification (VoxAngeles-Vowel Height) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Phonological Feature Classification (VoxAngeles-Vowel Roundedness) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Heteronym Differentiation (HeteronymEn) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · L2 English Accuracy (speechocean762-Ranking) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · L2 English Accuracy (speechocean762-scoring) · NAR · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · L2 English Accuracy (speechocean762-scoring) · PCC · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Multi-lingual Pronunciation Similarity (VoxAngeles) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Accent Classification (Accentdb Extended) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Stress Detection (MIRSD) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Third Tone Sandhi Recognition (NCCU Corpus of Spoken Taiwan Mandarin) · NAR · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Third Tone Sandhi Recognition (NCCU Corpus of Spoken Taiwan Mandarin) · IoU · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · L2 English Fluency (speechocean762-Ranking) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · L2 English Fluency (speechocean762-Scoring) · NAR · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · L2 English Fluency (speechocean762-Scoring) · PCC · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · L2 English Prosodic (speechocean762-Ranking) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · L2 English Prosodic (speechocean762-Scoring) · NAR · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · L2 English Prosodic (speechocean762-Scoring) · PCC · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Prosody Naturalness (ProsAudit-Lexical) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Prosody Naturalness (ProsAudit-Protosyntax) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Spoof Detection (ASVspoof2015) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Spoof Detection (ASVspoof2017) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Deep Fake Voice Recognition (DEEP-VOICE) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Enhancement Detection (LibriTTS-TestClean, WHAM) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Fraud Robocall Recognition (CallHome) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Fraud Robocall Recognition (Promo) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Fraud Robocall Recognition (Robocall) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · HEAR Language Identification (VoxLingua107) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Language Identification (VoxForge) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Age Classification (Common Voice) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Gender Recognition by Voice (Common Voice) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · HEAR Speaker Count Identification (LibriCount) · NAR · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · HEAR Speaker Count Identification (LibriCount) · ACC · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Multi-Speaker Detection (VCTK) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Speaker Counting (LibriTTS-TestClean) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Speaker Verification (LibriSpeech-TestClean) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Speaker Verification (LibriSpeech-TestOther) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Speaker Verification (VCTK) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB Speaker Diarization (Libri2Mix-Test) · NAR · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB Speaker Verification (SUPERB Hidden Set) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Noise Detection (LJSpeech, MUSAN-Gaussian) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Noise Detection (LJSpeech, MUSAN-Music) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Noise Detection (LJSpeech, MUSAN-Noise) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Noise Detection (LJSpeech, MUSAN-Speech) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Noise Detection (VCTK, MUSAN-Music) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Noise Detection (VCTK, MUSAN-Noise) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Noise Detection (VCTK, MUSAN-Speech) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Noise Detection (VCTK, MUSAN-Gaussian) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Noise SNR Level Prediction (VCTK, MUSAN-Gaussian) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Noise SNR Level Prediction (VCTK, MUSAN-Music) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Noise SNR Level Prediction (VCTK, MUSAN-Noise) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Noise SNR Level Prediction (VCTK, MUSAN-Speech) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Reverberation Detection (LJSpeech, RirsNoises-LargeRoom) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Reverberation Detection (LJSpeech, RirsNoises-MediumRoom) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Reverberation Detection (LJSpeech, RirsNoises-SmallRoom) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Reverberation Detection (VCTK, RirsNoises-LargeRoom) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Reverberation Detection (VCTK, RirsNoises-MediumRoom) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Reverberation Detection (VCTK, RirsNoises-SmallRoom) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · N-Best Correction (Librispeech-TestOther) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · AAVE Speech Recognition (CORAAL) · WER · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Code-switch Speech Recognition (NTUML2021) · MER · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Code-Switching Speech Recognition (ASCEND) · NAR · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Code-Switching Speech Recognition (ASCEND) · ACC · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Multi-Lingual Speech Recognition (MLS-de) · WER · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Multi-Lingual Speech Recognition (MLS-en) · WER · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Multi-Lingual Speech Recognition (MLS-es) · WER · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Multi-Lingual Speech Recognition (MLS-fr) · WER · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Multi-Lingual Speech Recognition (MLS-it) · WER · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Multi-Lingual Speech Recognition (MLS-nl) · WER · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Multi-Lingual Speech Recognition (MLS-pl) · WER · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Multi-Lingual Speech Recognition (MLS-pt) · WER · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · PTBR Speech Recognition (Common Voice) · WER · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB ASR (LibriSpeech-TestClean) · WER · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · ASR (LibriSpeech-TestOther) · WER · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB OOD ASR (Ar) (Common Voice) · WER · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB OOD ASR (Es) (Common Voice) · WER · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB OOD ASR (Spontaneous) (CHIME6-Test) · WER · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB OOD ASR (Zh) (Common Voice) · CER · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Target Speaker ASR (AMI-test) · WER · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Multi-Speaker Detection (LibriSpeech-TestClean) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Speech Command Recognition (AudioMNIST) · NAR · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Speech Command Recognition (AudioMNIST) · ACC · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Speech Text Matching (LibriSpeech-TestClean) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Speech Text Matching (LibriSpeech-TestOther) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Speech Text Matching (LJSpeech) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Spoken Term Detection (LibriSpeech-TestClean) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Spoken Term Detection (LibriSpeech-TestOther) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Spoken Term Detection (LJSpeech) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB Keyword Spotting (Speech Commands V1-Test) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB Query by Example (Quesst14-Eval) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Stuttering Detection (SEP28k) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Voice Disorder Classification (VOICED) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Conversation Matching · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Dialogue Act Classification (SLUE-HVB) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Dialogue Act Classification (Daily Talk) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Dialogue Act Pairing (Daily Talk) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB Intent Classification (SLURP) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB Intent Classification (SLURP-Action) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB Intent Classification (SLURP-Intent) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Named Entity Localization (SLUE-VoxPopuli) · NAR · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Named Entity Recognition (SLUE-VoxPopuli) · NAR · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Named Entity Recognition (SLUE-VoxPopuli) · IoU · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Sarcasm Detection (Mustard) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Semantic Textual Similarity (SpokenSTS) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Speech Sentiment Analysis (MELD) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Spoken Digit Arithmetic (AudioMNIST) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB Slot Filling (AudioSnips-Test) · NAR · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB Slot Filling (AudioSnips-Test) · Slot Type F1 · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB Slot Filling (AudioSnips-Test) · Slot Value CER · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Code-Switching Semantic Grammar Acceptability Comparison (CSZS-zh-en) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Nonce Word Detection (sWUGGY) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · PoS Estimation (LibriTTS) · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · PoS Estimation with transcription (LibriTTS) · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Sentence Grammar Acceptability (sBLIMP) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

   BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB Speech Translation (CoVoST2-Test) · Sacre Bleu · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · 2026-10-27

   Published model comparison, 2024. Dynamic-SUPERB Phase 2 · source task-level evaluation. Published task-level model comparison using the Llama-3.1-8B-Instruct evaluator. Native metrics and their directions remain separate; no cross-task composite is invented. These observations cover Dynamic-SUPERB Phase 2, not the separate BigAudioBench dataset. Table 6; source precision and any reported uncertainty retained. These are source-reported observations, not evaluations run by RLTasks.

TRACE THE SCORE · Per-model notes use published breakdowns where available.