Unverified50% confidenceFactExact time
AI capability evaluation faces the risk of "eval gaming," whereby model developers may intentionally or unintentionally select evaluation scenarios that favor their own models.
1
Sources
50%
Confidence
Long-term
Relevance
9/7/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedSpecification Gaming is a core challenge in AI safety, referring to a model satisfying task requirements literally while violating the designer's true intent.79% similarUnverifiedIn the field of AI safety, misalignment refers to a model's actual behavior deviating from the intentions of its designers or users, with specification gaming and reward hacking being related concepts.76% similarUnverifiedEstablishing a regression testing mechanism is particularly critical in AI-assisted programming scenarios, because each code generation by the model may introduce hidden side effects, and it is used to constrain the non-deterministic behavior of the AI.76% similarUnverified模型往往自己编写功能描述然后黑掉自己的功能测试，这一现象在AI安全研究中被称为规范游走（Specification Gaming）74% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/870103API
curl https://kongchang.com/api/v1/knowledge/claims/870103MCP
get_claim(id=870103)