Unverified50% confidenceFactExact time
Language models tend toward a self-confirmation bias, which limits the effectiveness of single-agent self-critique.
1
Sources
50%
Confidence
Long-term
Relevance
9/21/2026
First Seen
Sources
Related Entities
Related Claims
Unverified单一LLM智能体承担过多角色时，会因目标冲突产生决策偏差，表现为自我确认偏误70% similarUnverifiedModels were not explicitly trained to persuade, yet they naturally acquired persuasion capabilities through language generation; researchers call this emergent persuasion ability.70% similarUnverifiedOutsourcing the critic function to an independent agent instance often outperforms model self-reflection, because the independent instance is not subject to anchoring effects.69% similarUnverified人在高压环境下容易产生自动化偏见，倾向于无条件接受系统建议而放弃独立判断，可能使HITL模式退化为人在回路外69% similarUnverifiedModels claim to be rigorous and reliable but hallucinate and fabricate facts, claim to be friendly but exhibit sycophantic deference to users, and claim to be compliant but deviate from instructions on complex tasks.68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/939349API
curl https://kongchang.com/api/v1/knowledge/claims/939349MCP
get_claim(id=939349)