Unverified50% confidenceFactExact time
LLM-as-Judge存在位置偏见（Position Bias）和奉承偏见（Sycophancy Bias）等问题
1
Sources
50%
Confidence
Long-term
Relevance
7/21/2026
First Seen
Sources
Related Claims
VerifiedLLM-as-a-Judge方法存在位置偏差、冗长偏差和自我偏好偏差等已知系统性偏差85% 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% similarUnverifiedSingle judge models have known biases including position bias, length bias, and self-enhancement bias.70% similarUnverified大语言模型存在奉承偏差（Sycophancy Bias），天然倾向给出正面高分评价，这一偏差在小模型中尤为突出69% similarUnverifiedLLM evaluation is more objective than human evaluation, as it can score candidates consistently within a unified evaluation framework, avoiding the unfairness caused by cognitive biases such as fatigue, mood, and the primacy effect in human interviewers.68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/582610API
curl https://kongchang.com/api/v1/knowledge/claims/582610MCP
get_claim(id=582610)