Unverified50% confidenceOpinionExact time
Autoregressive language models have never seen 'I don't know' as a high-frequency reasonable output during training, so they are more likely to generate fabricated content at the boundaries of their knowledge than to express uncertainty
1
Sources
50%
Confidence
Long-term
Relevance
7/19/2026
First Seen
Sources
Related Claims
Unverified反事实问题能有效区分真正理解与模式匹配，因为这类问题几乎不可能直接出现在训练语料中72% similarUnverifiedWhen a model lacks accurate knowledge of a particular library, it tends to generate syntactically correct but semantically incorrect code—so-called 'confident errors.'72% similarUnverified对齐悖论指若模型真实理解比其能输出的语言更精细，则基于语言输出的对齐评估可能无法真正捕捉模型内部状态的对齐程度70% similarUnverifiedThe root cause of rising hallucination rates in reasoning models is misaligned incentives: reinforcement learning training causes models to tend to generate detailed and persuasive chains of thought, and human evaluators give higher scores to more elaborately argued answers, even when they contain factual errors.70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/561572API
curl https://kongchang.com/api/v1/knowledge/claims/561572MCP
get_claim(id=561572)