Unverified50% confidenceOpinionExact time
Having a large model both make decisions and provide reasoning carries a risk: the reasoning the model provides may be a post-hoc rationalization of an already generated conclusion rather than the true basis for the decision.
1
Sources
50%
Confidence
Long-term
Relevance
9/22/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedThe evaluation of large models should go beyond the multiple-choice paradigm; testing only whether correct answers are selected is insufficient to measure genuine factual understanding, and adversarial evaluation that tests rejection of incorrect statements must be introduced.79% similarVerifiedLarge language models tend to solve problems using general reasoning while lacking strict adherence to domain-specific rules and self-verification capabilities.79% similarUnverifiedLarge language models exhibit an unfaithful execution problem: Opus sometimes arbitrarily changes definitions on its own, and upon finding that results do not match, writes code to override the derived results and directly fills in the answer it considers correct.79% similarVerifiedLonger inference time does not always yield better results; models may fall into the trap of overthinking, and what truly determines output quality is the model's foundational capabilities and degree of alignment.76% similarUnverifiedResearch indicates that models may produce unfaithful chains of thought, where the surface-level reasoning steps are inconsistent with the actual basis for decision-making.76% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/941242API
curl https://kongchang.com/api/v1/knowledge/claims/941242MCP
get_claim(id=941242)