Unverified50% confidenceFactExact time
Many of the accepted responses already stated the violation, yet the model still gave a "resolved" status label.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/30/2026
First Seen
Valid until: 1/28/2027
Sources
Related Entities
Related Claims
UnverifiedDiagnostic counterexample feedback provides concrete evidence that directly locates problem boundaries, making it more effective than generic prompts or simply informing the agent that an error exists.69% similarUnverifiedThe system prompt should add 'cite only explicitly listed facts; if a certain fact is missing, say you will go confirm it' to fix the failure of answering undocumented questions.68% similarUnverifiedSchema-only post-hoc correction can achieve a number of safe completions comparable to verifier feedback; possible reasons are that the model's raw output already contains the correct content but in a messy format, or that verifier feedback mainly imposes format constraints rather than providing new reasoning signals.66% similarUnverifiedInference methods based on model internal features face a trade-off between false positives and false negatives65% similarUnverified「没有理由回滚」和「已确认改进」是两种不同状态，但在许多模型注册表中往往被记录为同一个上线状态，混淆会导致「置信度债务」65% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1114703API
curl https://kongchang.com/api/v1/knowledge/claims/1114703MCP
get_claim(id=1114703)