Unverified50% confidenceFactExact time
Most other frontier models on FinDendro likewise show a bias in which self-reports are systematically lower than actual performance.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
11/5/2026
First Seen
Valid until: 2/3/2027
Sources
Related Entities
Related Claims
UnverifiedResearch shows that a model's accuracy on self-criticism tasks is significantly lower than in scenarios where criticism is performed by another model.73% similarUnverifiedOn the strongest models, the differences between experimental groups fall within confidence intervals, and researchers characterized the results as directional patterns and soundness properties rather than proven performance lower bounds69% similarUnverifiedThe same model may rank very differently across different evaluation platforms.69% similarUnverifiedThere is no stable mapping between a model's self-evaluation score and its actual accuracy; rephrase the same prompt and the score may drift substantially.68% similarUnverifiedMedical fine-tuning changes the model's style of expressing itself, not its metacognition; after fine-tuning the model is less overconfident, but the internal gauge for distinguishing what it knows from what it does not has not become more accurate.67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1138978API
curl https://kongchang.com/api/v1/knowledge/claims/1138978MCP
get_claim(id=1138978)