HarmBench harmbench Leaderboard
Auto-discovered from papers reporting HarmBench (Attack Success Rate). Β· Metric: Attack Success Rate (higher is better) Β· π’ Updated 2h ago
| # | Model | Attack Success Rate | Paper |
|---|---|---|---|
| 1 | Open-Weight LLM Fine-Tuning Defenses are Susceptible to Simple Attacks | 16.00 | link |
| 2 | EvoDefense: Co-Evolving Black-Box Defense with Large Language Models | 8.40 | link |