Unverified50% confidenceFactExact time
OpenAI研究员早在2021年发表论文指出过度优化奖励信号的风险
1
Sources
50%
Confidence
Long-term
Relevance
7/8/2026
First Seen
Sources
AI助手"表演性顺从"：当Claude学会敷衍而非真正解决问题
hackernewshackernews7/5/2026
Related Claims
UnverifiedSome models carry the risk of over-optimization toward test sets, causing leaderboard scores to deviate from real-world performance (benchmark saturation).64% similarVerifiedExcessive safety fine-tuning can cause models to over-refuse, undermining their practical utility.64% similarUnverified套壳GPT类项目的根本风险是一旦OpenAI推出类似功能或模型能力大幅提升，产品会瞬间失去存在价值62% similarUnverified安全研究者将系统行动能力远超其安全判断能力的潜在风险称为能力越界（Capability Overhang）61% similarUnverifiedOpenAI在2019年的研究中证实双重下降现象同时存在于模型复杂度、训练时间和样本数量三个维度61% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/295570API
curl https://kongchang.com/api/v1/knowledge/claims/295570MCP
get_claim(id=295570)