Unverified50% confidenceBenchmarkExact time
The fine-tuned model learned to refuse (saying "not sure"), whereas the base model never said "not sure" in 85 opportunities.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/18/2026
First Seen
Valid until: 1/16/2027
Sources
Related Entities
Related Claims
UnverifiedThe model showed the highest confidence (95%) on underdetermined problems (problems with no unique correct answer), exhibiting a reverse calibration phenomenon.75% similarUnverifiedThere 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.74% 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% similarUnverifiedThe author claims that the off-the-shelf model they started with, faced with 85 hidden allergen samples, never output "uncertain".70% similarUnverifiedLarge language models do not inherently possess calibrated probability estimation capabilities, and the probability values output by a model do not represent actual occurrence rates.68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1060643API
curl https://kongchang.com/api/v1/knowledge/claims/1060643MCP
get_claim(id=1060643)