Unverified50% confidenceFactExact time
The confidence output of large language models is fundamentally a product of probability distributions rather than a genuine judgment of factual truth or falsehood, and may express high confidence in incorrect statements.
1
Sources
50%
Confidence
Long-term
Relevance
9/11/2026
First Seen
Sources
Related Entities
Related Claims
VerifiedThe core principle of large models is probabilistic text continuation, and the output is probabilistic sampling rather than deterministic output.78% similarUnverifiedThe probability figures provided by guided generative models are essentially 'well-constrained inferences' rather than directly verifiable truths.76% similarVerifiedLarge language models are fundamentally estimators of the conditional probability distribution P(output|input), with the goal of generating "statistically plausible" text rather than guaranteeing factual accuracy.76% similarVerified大语言模型的幻觉问题源于其本质是基于统计概率的序列预测系统，生成的Token基于条件概率最大化而非真正的逻辑推理75% similarVerified大模型幻觉的技术本质在于模型在训练时学习的是语言模式的概率分布而非事实本身74% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/899279API
curl https://kongchang.com/api/v1/knowledge/claims/899279MCP
get_claim(id=899279)