Unverified50% confidenceFactExact time
The first-run scorer only accepted lines beginning with `ANSWER:`, causing most answers judged as incorrect to actually be correct—for example, correct responses with the number and label on the same line, Markdown bold formatting, or LaTeX format were all scored 0.
1
Sources
50%
Confidence
Long-term
Relevance
10/11/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedTraditional end-to-end evaluation only looks at whether the final result passes or fails, and cannot distinguish whether the failure occurred at the retrieval stage or the editing stage.68% similarUnverifiedDeterministic scorers can reliably catch field format errors, but averaging them across the entire suite cannot tell you where the errors are concentrated65% similarUnverifiedThe vast majority of LLM evaluation tools only score answers after they have already been presented to the user, and cannot intercept erroneous outputs in real time.64% similarUnverifiedWhen validation fails, the system automatically retries, and when no answer can be found from the source material, it responds with "I don't know" rather than fabricating an answer.64% similarUnverifiedThe article's data and conclusions all come from a single UP host's demonstration, and certain competitor names in the benchmark tests contain descriptive inaccuracies.64% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/999498API
curl https://kongchang.com/api/v1/knowledge/claims/999498MCP
get_claim(id=999498)