Unverified50% confidenceCautionExact time
The facts (ground-truth answer points) used for evaluation must never be provided to the Agent being tested, otherwise the Agent will take shortcuts and cheat.
1
Sources
50%
Confidence
Long-term
Relevance
9/16/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedIf the agent fails to complete the task in both the benign and malicious versions, it indicates that its refusal was likely just confused by the injected text and not a genuine security judgment.73% similarUnverified干活的Agent绝不能给自己打分，评估器不应与执行任务的Agent共享同一上下文，否则会产生带偏见的评估结果71% similarUnverifiedWhen Agent results deviate, the completeness of the context should be examined rather than rushing to adjust the prompt.70% similarUnverified应移除无法带来可测量收益的Agent，要求每个Agent证明自己的存在价值67% similarUnverifiedCounterfactual evaluation faces a fundamental problem of causal inference: it is impossible to simultaneously observe the outcomes for the same individual under both intervention and non-intervention conditions.67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/925610API
curl https://kongchang.com/api/v1/knowledge/claims/925610MCP
get_claim(id=925610)