Unverified50% confidenceCautionExact time
The objectives learned by models during training often deviate from the values and behavioral norms that humans truly expect
1
Sources
50%
Confidence
Long-term
Relevance
9/5/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedGoodhart's Law states that when a metric becomes a target in itself it is no longer a good metric, and introducing test sets into training data causes a model's exam scores to decouple from its true capabilities77% 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.74% similarUnverified仅凭人类演示训练出的初始策略上限被人类玩家的认知偏见和习惯性操作模式所锁定，突破人类水平来自大规模自对弈72% similarUnverified在机器人学习研究中，训练失败时研究者通常首先怀疑模型架构或超参数选择，而数据质量问题往往是最后才被排查到的因素70% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/862076API
curl https://kongchang.com/api/v1/knowledge/claims/862076MCP
get_claim(id=862076)