PapersBuilders09:02Chinese content新框架显式校正 LLM 评审偏差,用更少对比恢复可靠排名做模型评测的朋友看看这篇:它把 LLM 当裁判的位置偏好、啰嗦偏好这些偏差直接建模掉,能省大量对比次数。#LLM-as-a-judge#benchmarks#latent variable modelsaarXiv cs.LG@Harshita Katoch et al. (4)SourceRead laterReadWorth followingUsefulFollow LLM-as-a-judge