Unverified50% confidenceFactExact time
LLMs' probability judgments drift under the influence of irrelevant factors such as how evidence is framed and the order in which it is presented.
1
Sources
50%
Confidence
Long-term
Relevance
9/11/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedAlthough LLMs do not directly output normalized probability vectors, decision confidence can be approximately estimated through various engineering methods.79% similarVerifiedLLMs generate output token by token in an autoregressive manner, with each word's selection based on probability sampling from the current context.75% similarUnverifiedLLM可通过分析输出token的对数概率分布、思维链末尾自我评估、温度采样统计答案分布熵值等手段近似估计决策置信度75% similarVerifiedThe root cause of LLM hallucination lies in the autoregressive generation mechanism—the model predicts the next Token based on a probability distribution rather than genuine logical reasoning.74% similarVerifiedThe essence of LLMs is statistical probability-based sequence prediction rather than true logical reasoning, which creates inherent limitations when handling long-context dependencies, global architectural consistency, and complex edge cases.74% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/897401API
curl https://kongchang.com/api/v1/knowledge/claims/897401MCP
get_claim(id=897401)