Unverified50% confidenceFactExact time
同一LLM在同一次推理中同时扮演创作者和批评者容易陷入自我确认偏误，倾向于认可自己生成的内容
1
Sources
50%
Confidence
Long-term
Relevance
7/16/2026
First Seen
Sources
Related Claims
UnverifiedThere is a self-evaluation bias problem when the same model both generates content and reviews it.74% similarUnverified单一LLM评判者存在位置偏见（倾向于给出现在前面的答案更高评分）和自我增强偏见（倾向于给自家风格的输出打高分）74% similarUnverified自我审查机制通常通过反思提示或独立的批评者模型实现，形成生成-评估-修正的闭环73% similarUnverifiedConfirmation bias refers to the tendency of people to seek, interpret, and remember information that is consistent with their existing beliefs while ignoring contradictory evidence.71% similarUnverifiedLLM在生成摘要时天然倾向于反映评论中的多数意见，混淆了信息频率与信息重要性71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/527971API
curl https://kongchang.com/api/v1/knowledge/claims/527971MCP
get_claim(id=527971)