Unverified85% confidenceFactTime unknown
Petri在AI安全技术体系中属于评估环节，不负责训练模型，而是在模型训练完成后系统性地检验对齐效果
1
Sources
85%
Confidence
Long-term
Relevance
5/31/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedPetri是Anthropic开发的一款开源工具，专门用于测试和评估AI模型的对齐表现80% similarUnverifiedPatronus AI focuses on safety and compliance evaluation, while Lakera specializes in detection and defense against attacks such as prompt injection65% similarUnverified理想的机器人策略评估流程应采用双阶段验证：仿真评估作初筛，真实世界物理测试完成最终验证63% similarUnverifiedCybersecurity evaluation refers to the process of systematically testing an AI model's capabilities in offensive and defensive cybersecurity scenarios, typically including tasks such as having the model attempt to discover vulnerabilities, write exploit code, and simulate social engineering attacks.62% similarUnverifiedThe PAO system labels simulation convergence points as positive samples and non-convergence points as negative samples, using a classification-based constraint handling method (such as Probability of Feasibility) to train a Gaussian process classifier to estimate feasibility probability.61% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/12274API
curl https://kongchang.com/api/v1/knowledge/claims/12274MCP
get_claim(id=12274)