ELSPR: Evaluator LLM Training Data Self-Purification on Non-Transitive Preferences via Tournament Graph Reconstruction
ELSPR: Evaluator LLM Training Data Self-Purification on Non-Transitive Preferences via Tournament Graph Reconstruction
Large language models (LLMs) are widely used as evaluators for open-ended tasks, while previous research has emphasized biases in LLM evaluations, the issue of non-transitivity in pairwise comparisons remains unresolved: non-transitive preferences for pairwise comparisons, where evaluators prefer A over B, B over C, but C over A. Our results suggest that low-quality training data may reduce the transitivity of preferences generated by the Evaluator LLM. To address this, We propose a graph-theoretic framework to analyze and mitigate this problem by modeling pairwise preferences as tournament graphs. We quantify non-transitivity and introduce directed graph structural entropy to measure the overall clarity of preferences. Our analysis reveals significant non-transitivity in advanced Evaluator LLMs (with Qwen2.5-Max exhibiting 67.96%), as well as high entropy values (0.8095 for Qwen2.5-Max), reflecting low overall clarity of preferences. To address this issue, we designed a filtering strategy, ELSPR, to eliminate preference data that induces non-transitivity, retaining only consistent and transitive preference data for model fine-tuning. Experiments demonstrate that models fine-tuned with filtered data reduce non-transitivity by 13.78% (from 64.28% to 50.50%), decrease structural entropy by 0.0879 (from 0.8113 to 0.7234), and align more closely with human evaluators (human agreement rate improves by 0.6% and Spearman correlation increases by 0.01).
Yan Yu、Yilun Liu、Minggui He、Shimin Tao、Weibin Meng、Xinhua Yang、Li Zhang、Hongxia Ma、Chang Su、Hao Yang、Fuliang Li
计算技术、计算机技术
Yan Yu,Yilun Liu,Minggui He,Shimin Tao,Weibin Meng,Xinhua Yang,Li Zhang,Hongxia Ma,Chang Su,Hao Yang,Fuliang Li.ELSPR: Evaluator LLM Training Data Self-Purification on Non-Transitive Preferences via Tournament Graph Reconstruction[EB/OL].(2025-05-23)[2025-06-09].https://arxiv.org/abs/2505.17691.点此复制
评论