PandaGuard is an open-source multi-agent framework and leaderboard developed by Beijing-AISI to systematically evaluate large language models’ jailbreak safety, together with the PandaBench benchmark that tests dozens of models across many attack and defense strategies.
Endorsements made here support Beijing Institute of AI Safety and Governance (Beijing-AISI).
PandaGuard is an open-source multi-agent framework and leaderboard developed by Beijing-AISI to systematically evaluate large language models’ jailbreak safety, together with the PandaBench benchmark that tests dozens of models across many attack and defense strategies.
Endorsements made here support Beijing Institute of AI Safety and Governance (Beijing-AISI).
Project Details
Updated 05/18/26 · By grantmaking.aiPandaGuard models LLM safety assessment as interactions between attacker, defender and judge agents, enabling reproducible evaluation of a wide range of jailbreak attack methods and defense mechanisms across different LLM back-ends. The associated PandaBench benchmark aggregates large-scale evaluation data on 40+ models, 19 types of attacks and 12 defense strategies, providing a standardized way to compare vulnerabilities, study trade-offs between safety and utility, and inform Beijing-AISI’s research on robust and governable language models.
People
Updated 05/18/26 · By grantmaking.aiPrincipal Investigator
Discussion
No comments yet. Be the first to share your thoughts.