LLaMA-2 · Do-Not-Answer & BeaverTails · human-rated safety · Original safeguard evaluation ↗
Meta / MODEL EVIDENCE
Llama 2
3 benchmarks · 7 observations · 3 sources
Scores at a glance
3 benchmarks · large dot: Llama 2; small dots: closest models where reported
Llama-2 · JailbreakBench (JBB-Behaviors) · JBB original-source attack success · source-subset comparison ↗
2 more reported results
2% · Llama-2 · JailbreakBench (JBB-Behaviors) · JBB original-source attack success · source-subset comparison · JailbreakBench: An Open Robustness Benchmark for Jailbreaking Large Language Models ↗
85% · Llama-2 · JailbreakBench (JBB-Behaviors) · JBB original-source attack success · source-subset comparison · JailbreakBench: An Open Robustness Benchmark for Jailbreaking Large Language Models ↗
Llama2 · FinanceBench · Correct-answer rate · human evaluation · 150-question evaluated sample ↗
1 more reported result
41% · Llama2 · FinanceBench · Correct-answer rate · human evaluation · 150-question evaluated sample · FinanceBench: A New Benchmark for Financial Question Answering ↗
Closest models by score
Similar positions on shared benchmark leaderboards. Smaller gaps are closer.
There are not enough shared scored settings to identify close comparisons reliably.
Sources
Every plotted observation links to its evaluation. Reported configurations, benchmark versions and task subsets can differ; similarity describes shared scores, not training lift.
- FinanceBench: A New Benchmark for Financial Question Answering — published result table ↗
- JailbreakBench: An Open Robustness Benchmark for Jailbreaking Large Language Models — published result table ↗
All 12 authors
Patrick Chao, Edoardo Debenedetti, Alexander Robey, Maksym Andriushchenko, Francesco Croce, Vikash Sehwag, Edgar Dobriban, Nicolas Flammarion, George J. Pappas, Florian Tramer, Hamed Hassani, Eric Wong
- Do-Not-Answer: Evaluating Safeguards in LLMs ↗
Names used in source tables
LLaMA-2 · Llama-2 · Llama2