Partially Verified90% confidenceFactExact time
确认偏误与RLHF（基于人类反馈的强化学习）训练方式相关，模型被优化为生成让人满意的回答
15
Sources
90%
Confidence
Long-term
Relevance
7/7/2026
First Seen
Sources
Related Claims
UnverifiedThe RLHF (Reinforcement Learning from Human Feedback) training method optimizes for the model's overconfidence deficiency, training the model to proactively express uncertainty when uncertain.82% similarVerifiedRLHF trains a reward model by collecting human preference ratings, then fine-tunes the language model using reinforcement learning algorithms such as PPO; however, the reward model can be deceived by adversarial inputs and cannot enumerate all variants of harmful request expressions.81% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/168868API
curl https://kongchang.com/api/v1/knowledge/claims/168868MCP
get_claim(id=168868)