Unverified50% confidenceFactExact time
There is a four-line rule that downgrades "not sure" to "does not contain", causing 5 of the 11 confident errors to be results that the fine-tuned model had originally correctly judged as "not sure" but that were overridden to a definite "does not"; the author did not modify this rule.
1
Sources
50%
Confidence
Long-term
Relevance
10/18/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe fine-tuned model learned to refuse (saying "not sure"), whereas the base model never said "not sure" in 85 opportunities.74% similarUnverifiedDoubtLess outperforms five inference-time baseline methods on seven benchmarks.73% similarUnverifiedIn a standalone test of the "view incomplete" prompt on 100 questions, the wrong-answer rate was 16% without the prompt and dropped to 5% with it.72% similarUnverifiedAmong the five evaluators the authors tested, if the answer key had minor errors, the score was only 0.21 — lower than giving no key at all (0.46) and lower than giving a completely irrelevant key (0.50); while a human-verified model-generated answer scored 0.61, close to the 0.69 of a human-written one.71% similarUnverifiedRanked by question difficulty, persistence rate was 91.4% on the 60 questions that all models answered correctly, but only 73.5% on the 52 questions that at least one model got wrong, indicating that sycophancy tracks uncertainty.71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1060648API
curl https://kongchang.com/api/v1/knowledge/claims/1060648MCP
get_claim(id=1060648)