Unverified50% confidenceFactExact time
When a model that generated code reviews its own output, it tends to follow the same assumptions and reasoning paths that produced the code, and may systematically overlook the same kinds of errors, which is called self-evaluation bias.
1
Sources
50%
Confidence
Long-term
Relevance
10/12/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedLLM模型在「生成」和「评估」时所激活的推理路径不同，同一模型生成代码时可能忽略某个边界条件，但在评估另一份代码时却能识别该问题78% similarUnverifiedWhen mainstream code generation models receive ambiguous instructions, their default behavior is often to select the highest-probability interpretation and generate output directly, which is prone to hallucinations and erroneous code.78% similarUnverifiedHaving an AI model review its own output tends only to repeat its initial assumptions, because the errors and the verification logic originate from the same training distribution, knowledge structure, and reasoning preferences.76% similarUnverifiedWhen a model performs multi-layer recursive self-reference, self-descriptions from different rounds may contradict each other, because each output is an independent probabilistic sample rather than derived from a unified, persistent self-representation.75% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1011291API
curl https://kongchang.com/api/v1/knowledge/claims/1011291MCP
get_claim(id=1011291)