LLM Robustness Leaderboard v1 --Technical report
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2025
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866915443258687488 |
|---|---|
| author | Lefebvre, Pierre Peigné - Feuillade-Montixi, Quentin David, Tom Miailhe, Nicolas |
| author_facet | Lefebvre, Pierre Peigné - Feuillade-Montixi, Quentin David, Tom Miailhe, Nicolas |
| contents | This technical report accompanies the LLM robustness leaderboard published by PRISM Eval for the Paris AI Action Summit. We introduce PRISM Eval Behavior Elicitation Tool (BET), an AI system performing automated red-teaming through Dynamic Adversarial Optimization that achieves 100% Attack Success Rate (ASR) against 37 of 41 state-of-the-art LLMs. Beyond binary success metrics, we propose a fine-grained robustness metric estimating the average number of attempts required to elicit harmful behaviors, revealing that attack difficulty varies by over 300-fold across models despite universal vulnerability. We introduce primitive-level vulnerability analysis to identify which jailbreaking techniques are most effective for specific hazard categories. Our collaborative evaluation with trusted third parties from the AI Safety Network demonstrates practical pathways for distributed robustness assessment across the community. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2508_06296 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | LLM Robustness Leaderboard v1 --Technical report Lefebvre, Pierre Peigné - Feuillade-Montixi, Quentin David, Tom Miailhe, Nicolas Artificial Intelligence Machine Learning This technical report accompanies the LLM robustness leaderboard published by PRISM Eval for the Paris AI Action Summit. We introduce PRISM Eval Behavior Elicitation Tool (BET), an AI system performing automated red-teaming through Dynamic Adversarial Optimization that achieves 100% Attack Success Rate (ASR) against 37 of 41 state-of-the-art LLMs. Beyond binary success metrics, we propose a fine-grained robustness metric estimating the average number of attempts required to elicit harmful behaviors, revealing that attack difficulty varies by over 300-fold across models despite universal vulnerability. We introduce primitive-level vulnerability analysis to identify which jailbreaking techniques are most effective for specific hazard categories. Our collaborative evaluation with trusted third parties from the AI Safety Network demonstrates practical pathways for distributed robustness assessment across the community. |
| title | LLM Robustness Leaderboard v1 --Technical report |
| topic | Artificial Intelligence Machine Learning |
| url | https://arxiv.org/abs/2508.06296 |