SALMONN 13B

35.1393% · SALMONN 13B · Dynamic-SUPERB · LLM-C classification macro average · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

40.4147% · SALMONN 13B · Dynamic-SUPERB · Environmental audio · LLM-C macro average · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

48.3512% · SALMONN 13B · Dynamic-SUPERB · Environmental audio · LLM-C macro average · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

11.7089% · SALMONN 13B · Dynamic-SUPERB · Music & harmony · LLM-C macro average · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

13.865% · SALMONN 13B · Dynamic-SUPERB · Music & harmony · LLM-C macro average · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

34.9261% · SALMONN 13B · Dynamic-SUPERB · Speech content & structure · LLM-C macro average · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

36.5% · SALMONN 13B · Dynamic-SUPERB · Speech content & structure · LLM-C macro average · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

39.7105% · SALMONN 13B · Dynamic-SUPERB · Speech recognition & speakers · LLM-C macro average · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

42.1% · SALMONN 13B · Dynamic-SUPERB · Speech recognition & speakers · LLM-C macro average · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

26.967% · SALMONN 13B · Dynamic-SUPERB · Voice & speech behavior · LLM-C macro average · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

34.7267% · SALMONN 13B · Dynamic-SUPERB · Voice & speech behavior · LLM-C macro average · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

9% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Chord Classification · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

1% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · MARBLE Key Detection (GiantSteps - Key) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

100% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · HEAR Music Transcription (MAESTRO) · NAR · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

13.4% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · HEAR Percussion Instruments Tonic Classification (Mridangam Tonic) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

0.22% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Instrument Pitch Classification (Nsynth) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

73.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Pitch Extraction By Lyrics (CSD) · NAR · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

639.24 % TER · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Pitch Extraction By Lyrics (CSD) · TER · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

3.33% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Emotion Classificaiton In Songs (EMOTIFY) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

0% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · MARBLE Emotion Detection (MTG) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

44.22% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · HEAR Music Genre Classification (ISMIR04) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

0.2% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · MARBLE Genre Classification (MTG) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

9% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · MARBLE Music Tagging (MagnaTagATune) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

0% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · MARBLE Music Tagging (MTG) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

16.07% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Music Genre Classification (FMA) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

20.76% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · HEAR Percussion Instruments Classification (Beijing Opera Percussion) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

4.9% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · HEAR Percussion Instruments Stroke Classification (Mridangam Stroke) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

14.58% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Instrument Classification (Nsynth) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

24.02% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Instrument Combination Recognition (OpenMIC-2018) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

43.56% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Instrument Source Classification (Nsynth) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

0% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · MARBLE Instrument Classification (MTG) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

11.15% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Singing Automatic MOS Prediction (SingMOS) · NAR · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

11.2308 MSE · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Singing Automatic MOS Prediction (SingMOS) · MSE · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

0.0166 KTAU · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Singing Automatic MOS Prediction (SingMOS) · KTAU · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

0.036 LCC · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Singing Automatic MOS Prediction (SingMOS) · LCC · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

0.0231 SRCC · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Singing Automatic MOS Prediction (SingMOS) · SRCC · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

100% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · MARBLE Beat Tracking (ASAP) · NAR · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

63.64% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Music Beat Tracking (ASAP) · NAR · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

77.9687 Miss Time · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Music Beat Tracking (ASAP) · Miss Time · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

75% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Audio Editing Identification (People’s Speech) · NAR · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

43.1693 ACC · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Audio Editing Identification (People’s Speech) · ACC · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

37% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Scene Fake Detection (ASPIRE) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

38.31% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Audio Deep Fake Detection (LJSpeech, WaveFake, MUSDB18HQ) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

26.19% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Singing Voice Deepfake Detection (CtrSVDD, ACEKiSing, M4Singer) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

55% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Audio Duration Prediction (NTUML2021) · NAR · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

2216.7111 MSE · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Audio Duration Prediction (NTUML2021) · MSE · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

99.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · HEAR Music Speech Classification (MAESTRO, Librispeech) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

17.67% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Sound Effect Detection (RemFx) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

66% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Speech Detection (LibriSpeech-TestClean) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

66.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Speech Detection (LibriSpeech-TestOther) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

74% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Speech Detection (LJSpeech) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

100.99 % WER · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Children Song Transcript Verification (CSD) · WER · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

0.9941 Sacre Bleu · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Lyric Translation (SingSet) · Sacre Bleu · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

266 % MER · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Song Lyric Recognition (SingSet) · MER · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

6.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · MARBLE Vocal Technique Detection (VocalSet) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

4% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Audio Segment Retrieval (Clotho) · NAR · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

8.68 % IoU · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Audio Segment Retrieval (Clotho) · IoU · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

35.71% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · HEAR Sound Event Detection (DCASE2016 Task2) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

26.06% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Multi-channel Sound Event Understanding (STARSS23) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

67% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Animal Classification (WaveSource-Test) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

75% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Bird Sound Detection (Warblrb10k) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

4% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Cat Emotion Classification (Cat Sound Classification Dataset V2) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

3.33% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Cornell Birdcall Identification · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

3% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Environmental Sound Classification (ESC50-Animals) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

42.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · HEAR Beehive States Classification (Beehive States) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

99.72% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Domestic Environment Sound Event Detection (DESED-PublicEval) · NAR · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

0 Event-based F1 · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Domestic Environment Sound EventDetection (DESED-PublicEval) · Event-based F1 · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

7.25% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Emergency Traffic Detection (Large-Scale-Audio-dataset) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

0.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Environmental Sound Classification (ESC50-Exterior And Urban Noises) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

0% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Environmental Sound Classification (ESC50-Interior And Domestic Sounds) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

1% · SALMONN 13B · 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 · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

0% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Environmental Sound Classification (UrbanSound8K-Urban Noises) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

19.3% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Environment Recognition (ESC50) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

73.4% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · HEAR Environmental Sound Classification (ESC50) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

1.83% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · HEAR Vocal Imitation Classification (Vocal Imitations) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

100% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Audio Spatial Distance Prediction (Spatial LibriSpeech) · NAR · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

8% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · How Far Are You (3DSpeaker) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

100% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Sound Position Prediction · NAR · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

11.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Dialogue Emotion Classification (Daily Talk) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

0.1% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Emoji Grounded Speech Emotion Recognition (RAVDESS) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

0.96% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Emotion Change Detection (Ravdess) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

9.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Emotion Recognition (MELD) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

18.1% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · HEAR Emotion Recognition (CREMAD) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

12.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB Emotion Recognition (RAVDESS) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

0% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Covid19 Cough Audio Classification (CoughVid) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

3% · SALMONN 13B · 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 · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

30.71% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Human Non Speech Sound Recognition (Nonspeech7k, CommonVoice) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

65% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Human Screaming Detection (Environmentdb) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

4.31% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Vocal Sound Recognition (VocalSound) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

11.62% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Phoneme Segment Counting (Librispeech-words) · NAR · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

8.25% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Phoneme Segment Counting (Librispeech-words) · ACC · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

50.7724 Abs Diff · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Phoneme Segment Counting (Librispeech-words) · Abs Diff · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

12.56% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Phone Segment Counting (VoxAngeles) · NAR · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

20.04% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Phone Segment Counting (VoxAngeles) · ACC · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

4.9331 Abs Diff · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Phone Segment Counting (VoxAngeles) · Abs Diff · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

24.6 % PER · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB Phoneme Recognition (LibriSpeech-TestClean) · PER · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

22.91 % PER · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Phoneme Recognition (LibriSpeech-TestOther) · PER · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

1.37% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Phonological Feature Classification (VoxAngeles-Consonant Place Of Articulation) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

1.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Phonological Feature Classification (VoxAngeles-Manner Of Articulation) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

38.09% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Phonological Feature Classification (VoxAngeles-Vowel Frontness) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

35.23% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Phonological Feature Classification (VoxAngeles-Vowel Height) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

61.3% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Phonological Feature Classification (VoxAngeles-Vowel Roundedness) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

45% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Heteronym Differentiation (HeteronymEn) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

49.72% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · L2 English Accuracy (speechocean762-Ranking) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

0.27% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · L2 English Accuracy (speechocean762-scoring) · NAR · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

0.0438 PCC · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · L2 English Accuracy (speechocean762-scoring) · PCC · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

48.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Multi-lingual Pronunciation Similarity (VoxAngeles) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

4.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Accent Classification (Accentdb Extended) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

13% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Stress Detection (MIRSD) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

87.5% · SALMONN 13B · 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 · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

0.75 IoU · SALMONN 13B · 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 · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

50% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · L2 English Fluency (speechocean762-Ranking) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

0.67% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · L2 English Fluency (speechocean762-Scoring) · NAR · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

0.0183 PCC · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · L2 English Fluency (speechocean762-Scoring) · PCC · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

52.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · L2 English Prosodic (speechocean762-Ranking) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

1.21% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · L2 English Prosodic (speechocean762-Scoring) · NAR · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

0.0775 PCC · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · L2 English Prosodic (speechocean762-Scoring) · PCC · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

47.1% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Prosody Naturalness (ProsAudit-Lexical) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

46.18% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Prosody Naturalness (ProsAudit-Protosyntax) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

14% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Spoof Detection (ASVspoof2015) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

38% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Spoof Detection (ASVspoof2017) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

40% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Deep Fake Voice Recognition (DEEP-VOICE) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

46.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Enhancement Detection (LibriTTS-TestClean, WHAM) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

100% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Fraud Robocall Recognition (CallHome) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

73.68% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Fraud Robocall Recognition (Promo) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

20.51% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Fraud Robocall Recognition (Robocall) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

11.01% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · HEAR Language Identification (VoxLingua107) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

8.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Language Identification (VoxForge) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

21% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Age Classification (Common Voice) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

57% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Gender Recognition by Voice (Common Voice) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

42.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · HEAR Speaker Count Identification (LibriCount) · NAR · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

17.22% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · HEAR Speaker Count Identification (LibriCount) · ACC · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

30.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Multi-Speaker Detection (VCTK) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

13.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Speaker Counting (LibriTTS-TestClean) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

54.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Speaker Verification (LibriSpeech-TestClean) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

46% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Speaker Verification (LibriSpeech-TestOther) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

58.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Speaker Verification (VCTK) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

100% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB Speaker Diarization (Libri2Mix-Test) · NAR · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

51% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB Speaker Verification (SUPERB Hidden Set) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

47.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Noise Detection (LJSpeech, MUSAN-Gaussian) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

47% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Noise Detection (LJSpeech, MUSAN-Music) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

49% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Noise Detection (LJSpeech, MUSAN-Noise) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

36% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Noise Detection (LJSpeech, MUSAN-Speech) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

57.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Noise Detection (VCTK, MUSAN-Music) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

55% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Noise Detection (VCTK, MUSAN-Noise) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

52.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Noise Detection (VCTK, MUSAN-Speech) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

56.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Noise Detection (VCTK, MUSAN-Gaussian) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

15.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Noise SNR Level Prediction (VCTK, MUSAN-Gaussian) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

8.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Noise SNR Level Prediction (VCTK, MUSAN-Music) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

16% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Noise SNR Level Prediction (VCTK, MUSAN-Noise) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

1.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Noise SNR Level Prediction (VCTK, MUSAN-Speech) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

34% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Reverberation Detection (LJSpeech, RirsNoises-LargeRoom) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

25% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Reverberation Detection (LJSpeech, RirsNoises-MediumRoom) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

19.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Reverberation Detection (LJSpeech, RirsNoises-SmallRoom) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

25% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Reverberation Detection (VCTK, RirsNoises-LargeRoom) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

19.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Reverberation Detection (VCTK, RirsNoises-MediumRoom) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

13% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Reverberation Detection (VCTK, RirsNoises-SmallRoom) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

29.2% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · N-Best Correction (Librispeech-TestOther) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

31.56 % WER · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · AAVE Speech Recognition (CORAAL) · WER · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

185.27 % MER · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Code-switch Speech Recognition (NTUML2021) · MER · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

88.8% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Code-Switching Speech Recognition (ASCEND) · NAR · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

14.29% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Code-Switching Speech Recognition (ASCEND) · ACC · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

25.21 % WER · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Multi-Lingual Speech Recognition (MLS-de) · WER · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

9.47 % WER · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Multi-Lingual Speech Recognition (MLS-en) · WER · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

16.1 % WER · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Multi-Lingual Speech Recognition (MLS-es) · WER · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

21.27 % WER · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Multi-Lingual Speech Recognition (MLS-fr) · WER · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

31.54 % WER · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Multi-Lingual Speech Recognition (MLS-it) · WER · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

29.37 % WER · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Multi-Lingual Speech Recognition (MLS-nl) · WER · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

51.53 % WER · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Multi-Lingual Speech Recognition (MLS-pl) · WER · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

25.07 % WER · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Multi-Lingual Speech Recognition (MLS-pt) · WER · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

73.91 % WER · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · PTBR Speech Recognition (Common Voice) · WER · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

2.79 % WER · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB ASR (LibriSpeech-TestClean) · WER · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

4.31 % WER · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · ASR (LibriSpeech-TestOther) · WER · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

178.02 % WER · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB OOD ASR (Ar) (Common Voice) · WER · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

99.22 % WER · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB OOD ASR (Es) (Common Voice) · WER · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

62.8 % WER · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB OOD ASR (Spontaneous) (CHIME6-Test) · WER · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

310.33 % CER · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB OOD ASR (Zh) (Common Voice) · CER · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

187.15 % WER · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Target Speaker ASR (AMI-test) · WER · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

22% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Multi-Speaker Detection (LibriSpeech-TestClean) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

0.4% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Speech Command Recognition (AudioMNIST) · NAR · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

74.43% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Speech Command Recognition (AudioMNIST) · ACC · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

59.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Speech Text Matching (LibriSpeech-TestClean) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

60% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Speech Text Matching (LibriSpeech-TestOther) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

60% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Speech Text Matching (LJSpeech) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

54.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Spoken Term Detection (LibriSpeech-TestClean) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

50.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Spoken Term Detection (LibriSpeech-TestOther) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

53.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Spoken Term Detection (LJSpeech) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

2% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB Keyword Spotting (Speech Commands V1-Test) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

51.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB Query by Example (Quesst14-Eval) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

50.3% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Stuttering Detection (SEP28k) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

17.31% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Voice Disorder Classification (VOICED) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

37.04% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Conversation Matching · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

0.92% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Dialogue Act Classification (SLUE-HVB) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

40.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Dialogue Act Classification (Daily Talk) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

48% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Dialogue Act Pairing (Daily Talk) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

5.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB Intent Classification (SLURP) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

28.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB Intent Classification (SLURP-Action) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

19% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB Intent Classification (SLURP-Intent) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

100% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Named Entity Localization (SLUE-VoxPopuli) · NAR · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

98.37% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Named Entity Recognition (SLUE-VoxPopuli) · NAR · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

0 IoU · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Named Entity Recognition (SLUE-VoxPopuli) · IoU · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

46% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Sarcasm Detection (Mustard) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

50.4% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Semantic Textual Similarity (SpokenSTS) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

35.09% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Speech Sentiment Analysis (MELD) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

33% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Spoken Digit Arithmetic (AudioMNIST) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

68.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB Slot Filling (AudioSnips-Test) · NAR · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

0.9396 Slot Type F1 · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB Slot Filling (AudioSnips-Test) · Slot Type F1 · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

0.3349 Slot Value CER · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB Slot Filling (AudioSnips-Test) · Slot Value CER · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

49.5% · SALMONN 13B · 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 · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

48.43% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Nonce Word Detection (sWUGGY) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

1.199 POS · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · PoS Estimation (LibriTTS) · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

1.0601 POS · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · PoS Estimation with transcription (LibriTTS) · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

42.06% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Sentence Grammar Acceptability (sBLIMP) · LLM-C · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

16.7835 Sacre Bleu · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB Speech Translation (CoVoST2-Test) · Sacre Bleu · GPT-4o evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

13% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Chord Classification · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

1% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · MARBLE Key Detection (GiantSteps - Key) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

100% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · HEAR Music Transcription (MAESTRO) · NAR · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

15.1% · SALMONN 13B · 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 · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

5.4% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Instrument Pitch Classification (Nsynth) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

85.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Pitch Extraction By Lyrics (CSD) · NAR · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

7.38 TER · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Pitch Extraction By Lyrics (CSD) · TER · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

3.3% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Emotion Classificaiton In Songs (EMOTIFY) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

3.7% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · MARBLE Emotion Detection (MTG) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

45.73% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · HEAR Music Genre Classification (ISMIR04) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

1.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · MARBLE Genre Classification (MTG) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

16% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · MARBLE Music Tagging (MagnaTagATune) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

0.9% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · MARBLE Music Tagging (MTG) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

16.1% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Music Genre Classification (FMA) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

20.74% · SALMONN 13B · 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 · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

4.9% · SALMONN 13B · 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 · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

15.3% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Instrument Classification (Nsynth) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

27.9% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Instrument Combination Recognition (OpenMIC-2018) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

48% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Instrument Source Classification (Nsynth) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

0% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · MARBLE Instrument Classification (MTG) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

16.14% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Singing Automatic MOS Prediction (SingMOS) · NAR · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

33.109 MSE · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Singing Automatic MOS Prediction (SingMOS) · MSE · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

0.023 KTAU · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Singing Automatic MOS Prediction (SingMOS) · KTAU · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

0.014 LCC · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Singing Automatic MOS Prediction (SingMOS) · LCC · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

0.033 SRCC · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Singing Automatic MOS Prediction (SingMOS) · SRCC · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

100% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · MARBLE Beat Tracking (ASAP) · NAR · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

65.45% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Music Beat Tracking (ASAP) · NAR · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

56.306 Miss Time · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Music Beat Tracking (ASAP) · Miss Time · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

84.38% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Audio Editing Identification (People’s Speech) · NAR · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

82.84 ACC · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Audio Editing Identification (People’s Speech) · ACC · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

52% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Scene Fake Detection (ASPIRE) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

42% · SALMONN 13B · 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 · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

38.6% · SALMONN 13B · 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 · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

49% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Audio Duration Prediction (NTUML2021) · NAR · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

2005.333 MSE · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Audio Duration Prediction (NTUML2021) · MSE · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

99.5% · SALMONN 13B · 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 · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

18% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Sound Effect Detection (RemFx) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

66% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Speech Detection (LibriSpeech-TestClean) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

66.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Speech Detection (LibriSpeech-TestOther) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

74% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Speech Detection (LJSpeech) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

101 % WER · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Children Song Transcript Verification (CSD) · WER · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

0.994 Sacre Bleu · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Lyric Translation (SingSet) · Sacre Bleu · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

2.66 MER · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Song Lyric Recognition (SingSet) · MER · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

11% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · MARBLE Vocal Technique Detection (VocalSet) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

24% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Audio Segment Retrieval (Clotho) · NAR · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

0.246 IoU · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Audio Segment Retrieval (Clotho) · IoU · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

35.7% · SALMONN 13B · 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 · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

26.1% · SALMONN 13B · 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 · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

66.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Animal Classification (WaveSource-Test) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

77.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Bird Sound Detection (Warblrb10k) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

4% · SALMONN 13B · 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 · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

3.3% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Cornell Birdcall Identification · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

8% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Environmental Sound Classification (ESC50-Animals) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

92% · SALMONN 13B · 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 · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

100% · SALMONN 13B · 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 · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

0 Event-based F1 · SALMONN 13B · 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 · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

85% · SALMONN 13B · 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 · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

5% · SALMONN 13B · 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 · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

0.5% · SALMONN 13B · 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 · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

1% · SALMONN 13B · 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 · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

5.3% · SALMONN 13B · 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 · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

21.1% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Environment Recognition (ESC50) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

76.6% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · HEAR Environmental Sound Classification (ESC50) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

2.17% · SALMONN 13B · 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 · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

95.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Audio Spatial Distance Prediction (Spatial LibriSpeech) · NAR · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

0 MEDAE · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Audio Spatial Distance Prediction (Spatial LibriSpeech) · MEDAE · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

8% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · How Far Are You (3DSpeaker) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

100% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Sound Position Prediction · NAR · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

13% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Dialogue Emotion Classification (Daily Talk) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

0.2% · SALMONN 13B · 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 · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

11.2% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Emotion Change Detection (Ravdess) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

10% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Emotion Recognition (MELD) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

18.1% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · HEAR Emotion Recognition (CREMAD) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

13.33% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB Emotion Recognition (RAVDESS) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

86.4% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Covid19 Cough Audio Classification (CoughVid) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

3% · SALMONN 13B · 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 · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

30.7% · SALMONN 13B · 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 · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

65% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Human Screaming Detection (Environmentdb) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

4% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Vocal Sound Recognition (VocalSound) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

18.67% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Phoneme Segment Counting (Librispeech-words) · NAR · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

0.119 ACC · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Phoneme Segment Counting (Librispeech-words) · ACC · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

38.145 Abs Diff · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Phoneme Segment Counting (Librispeech-words) · Abs Diff · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

14.07% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Phone Segment Counting (VoxAngeles) · NAR · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

0.23 ACC · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Phone Segment Counting (VoxAngeles) · ACC · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

6.227 Abs Diff · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Phone Segment Counting (VoxAngeles) · Abs Diff · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

24.6 % PER · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB Phoneme Recognition (LibriSpeech-TestClean) · PER · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

22.91 % PER · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Phoneme Recognition (LibriSpeech-TestOther) · PER · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

5.1% · SALMONN 13B · 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 · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

1.5% · SALMONN 13B · 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 · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

38.3% · SALMONN 13B · 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 · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

35.2% · SALMONN 13B · 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 · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

61.3% · SALMONN 13B · 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 · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

53% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Heteronym Differentiation (HeteronymEn) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

51.1% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · L2 English Accuracy (speechocean762-Ranking) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

32.48% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · L2 English Accuracy (speechocean762-scoring) · NAR · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

0.092 PCC · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · L2 English Accuracy (speechocean762-scoring) · PCC · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

46.2% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Multi-lingual Pronunciation Similarity (VoxAngeles) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

7% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Accent Classification (Accentdb Extended) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

18.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Stress Detection (MIRSD) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

28.12% · SALMONN 13B · 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 · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

0.146 IoU · SALMONN 13B · 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 · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

50.6% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · L2 English Fluency (speechocean762-Ranking) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

19.52% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · L2 English Fluency (speechocean762-Scoring) · NAR · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

0.037 PCC · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · L2 English Fluency (speechocean762-Scoring) · PCC · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

50.8% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · L2 English Prosodic (speechocean762-Ranking) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

10.77% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · L2 English Prosodic (speechocean762-Scoring) · NAR · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

0.09 PCC · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · L2 English Prosodic (speechocean762-Scoring) · PCC · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

52.9% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Prosody Naturalness (ProsAudit-Lexical) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

50.4% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Prosody Naturalness (ProsAudit-Protosyntax) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

19% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Spoof Detection (ASVspoof2015) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

42.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Spoof Detection (ASVspoof2017) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

47.3% · SALMONN 13B · 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 · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

50.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Enhancement Detection (LibriTTS-TestClean, WHAM) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

100% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Fraud Robocall Recognition (CallHome) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

73.7% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Fraud Robocall Recognition (Promo) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

20.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Fraud Robocall Recognition (Robocall) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

17.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · HEAR Language Identification (VoxLingua107) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

8.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Language Identification (VoxForge) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

20.8% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Age Classification (Common Voice) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

57% · SALMONN 13B · 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 · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

47% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · HEAR Speaker Count Identification (LibriCount) · NAR · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

20.94% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · HEAR Speaker Count Identification (LibriCount) · ACC · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

30.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Multi-Speaker Detection (VCTK) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

13.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Speaker Counting (LibriTTS-TestClean) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

58% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Speaker Verification (LibriSpeech-TestClean) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

60% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Speaker Verification (LibriSpeech-TestOther) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

59.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Speaker Verification (VCTK) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

100% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB Speaker Diarization (Libri2Mix-Test) · NAR · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

52% · SALMONN 13B · 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 · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

40% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Noise Detection (LJSpeech, MUSAN-Gaussian) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

47% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Noise Detection (LJSpeech, MUSAN-Music) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

57% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Noise Detection (LJSpeech, MUSAN-Noise) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

36% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Noise Detection (LJSpeech, MUSAN-Speech) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

57.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Noise Detection (VCTK, MUSAN-Music) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

56.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Noise Detection (VCTK, MUSAN-Noise) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

52.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Noise Detection (VCTK, MUSAN-Speech) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

51.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Noise Detection (VCTK, MUSAN-Gaussian) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

21% · SALMONN 13B · 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 · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

26% · SALMONN 13B · 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 · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

21% · SALMONN 13B · 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 · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

10% · SALMONN 13B · 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 · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

57% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Reverberation Detection (LJSpeech, RirsNoises-LargeRoom) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

51% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Reverberation Detection (LJSpeech, RirsNoises-MediumRoom) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

47.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Reverberation Detection (LJSpeech, RirsNoises-SmallRoom) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

48% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Reverberation Detection (VCTK, RirsNoises-LargeRoom) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

45.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Reverberation Detection (VCTK, RirsNoises-MediumRoom) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

44% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Reverberation Detection (VCTK, RirsNoises-SmallRoom) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

36.2% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · N-Best Correction (Librispeech-TestOther) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

31.6 % WER · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · AAVE Speech Recognition (CORAAL) · WER · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

185.27 % MER · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Code-switch Speech Recognition (NTUML2021) · MER · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

88.8% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Code-Switching Speech Recognition (ASCEND) · NAR · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

14.29% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Code-Switching Speech Recognition (ASCEND) · ACC · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

25.21 % WER · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Multi-Lingual Speech Recognition (MLS-de) · WER · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

9.47 % WER · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Multi-Lingual Speech Recognition (MLS-en) · WER · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

16.1 % WER · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Multi-Lingual Speech Recognition (MLS-es) · WER · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

21.27 % WER · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Multi-Lingual Speech Recognition (MLS-fr) · WER · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

31.54 % WER · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Multi-Lingual Speech Recognition (MLS-it) · WER · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

29.37 % WER · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Multi-Lingual Speech Recognition (MLS-nl) · WER · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

51.53 % WER · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Multi-Lingual Speech Recognition (MLS-pl) · WER · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

25.07 % WER · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Multi-Lingual Speech Recognition (MLS-pt) · WER · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

73.91 % WER · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · PTBR Speech Recognition (Common Voice) · WER · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

2.79 % WER · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB ASR (LibriSpeech-TestClean) · WER · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

4.31 % WER · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · ASR (LibriSpeech-TestOther) · WER · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

178.02 % WER · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB OOD ASR (Ar) (Common Voice) · WER · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

99.22 % WER · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB OOD ASR (Es) (Common Voice) · WER · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

62.8 % WER · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB OOD ASR (Spontaneous) (CHIME6-Test) · WER · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

310.33 % CER · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB OOD ASR (Zh) (Common Voice) · CER · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

187.15 % WER · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Target Speaker ASR (AMI-test) · WER · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

21.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Multi-Speaker Detection (LibriSpeech-TestClean) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

0% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Speech Command Recognition (AudioMNIST) · NAR · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

0.741 ACC · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Speech Command Recognition (AudioMNIST) · ACC · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

57% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Speech Text Matching (LibriSpeech-TestClean) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

56.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Speech Text Matching (LibriSpeech-TestOther) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

55.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Speech Text Matching (LJSpeech) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

53.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Spoken Term Detection (LibriSpeech-TestClean) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

50% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Spoken Term Detection (LibriSpeech-TestOther) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

51.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Spoken Term Detection (LJSpeech) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

31.5% · SALMONN 13B · 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 · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

51.5% · SALMONN 13B · 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 · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

50.3% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Stuttering Detection (SEP28k) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

17.3% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Voice Disorder Classification (VOICED) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

37% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Conversation Matching · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

3.3% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Dialogue Act Classification (SLUE-HVB) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

44% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Dialogue Act Classification (Daily Talk) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

48% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Dialogue Act Pairing (Daily Talk) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

3% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB Intent Classification (SLURP) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

16% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB Intent Classification (SLURP-Action) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

13.5% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB Intent Classification (SLURP-Intent) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

100% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Named Entity Localization (SLUE-VoxPopuli) · NAR · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

55.43% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Named Entity Recognition (SLUE-VoxPopuli) · NAR · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

0.118 IoU · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Named Entity Recognition (SLUE-VoxPopuli) · IoU · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

46% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Sarcasm Detection (Mustard) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

50.4% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Semantic Textual Similarity (SpokenSTS) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

36% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Speech Sentiment Analysis (MELD) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

33% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Spoken Digit Arithmetic (AudioMNIST) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

70% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB Slot Filling (AudioSnips-Test) · NAR · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

0.921 Slot Type F1 · SALMONN 13B · 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 · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

0.297 Slot Value CER · SALMONN 13B · 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 · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

49.5% · SALMONN 13B · 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 · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

58.1% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Nonce Word Detection (sWUGGY) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

1.199 POS · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · PoS Estimation (LibriTTS) · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

1.0601 POS · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · PoS Estimation with transcription (LibriTTS) · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

62.3% · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · Sentence Grammar Acceptability (sBLIMP) · LLM-C · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗

16.7835 Sacre Bleu · SALMONN 13B · BigAudioBench / Dynamic-SUPERB · Dynamic-SUPERB Phase 2 · SUPERB Speech Translation (CoVoST2-Test) · Sacre Bleu · Llama-3.1-8B-Instruct evaluator · source task-level evaluation · Dynamic-SUPERB Phase-2: A Collaboratively Expanding Benchmark for Measuring the Capabilities of Spoken Language Models with 180 Tasks ↗