1
Am I More Pointwise or Pairwise? Revealing Position Bias in Rubric-Based LLM-as-a-Judge
揭秘LLM作为评判者时的位置偏差，点式与对式评估究竟何者更准确？
arXiv:2602.02219v2 Announce Type: replace Abstract: Large language models are widely employed as evaluators, a paradigm commonly referred to as LLM-as…