用比赛淘汰赛方式让大模型更准地排序文档，提升复杂检索效果。
BracketRank: Large Language Model Document Ranking via Reasoning-based Competitive Elimination
- 把文档排序设计成带推理的淘汰赛，分组和流程自适应上下文限制。
- 在BRIGHT上达26.56 nDCG@10，远超RankGPT-4（17.0）等基线。
- 适合需要深度推理的复杂检索任务，如问答系统与知识挖掘。
深度语义推理的检索要求超越表层关键词匹配，而现有基于大模型的重排序器受限于上下文长度和顺序敏感性。我们提出 extbf{racketrank}，将文档重排序视为一种基于推理的竞赛淘汰机制。该方法包含三项创新：(1) 根据模型上下文限制自适应分组；(2) 使用强化推理的提示，要求逐步解释相关性；(3) 采用锦标赛结构，含胜者组与败者组。该设计保障了文档稳健推进，同时支持多阶段并行处理。在BRIGHT推理基准上的评估显示，racketrank 达到26.56 nDCG@10，显著优于最先进基线（如RankGPT-4为17.0，Rank-R1-14B为20.5）。在TREC数据集上，racketrank 在DL 19上取得77.90 nDCG@5，DL 20上为75.85 nDCG@5，均超过所有基线，证明在竞争淘汰中引入显式推理是解决复杂多步检索任务的有效范式。
原文摘要 · Abstract (English)
Reasoning-intensive retrieval requires deep semantic inference beyond surface-level keyword matching, posing a challenge for current LLM-based rerankers limited by context constraints and order sensitivity. We propose \textbf{\BracketRank}, a framework that treats document reranking as a reasoning-driven competitive tournament. Our approach introduces three key innovations: (1) adaptive grouping based on model context limits, (2) reasoning-enhanced prompts that mandate step-by-step relevance explanations, and (3) a bracket-style elimination structure with winner and loser tracks. This design ensures robust document advancement while enabling parallel processing across competition stages. Evaluation on the BRIGHT reasoning benchmark shows that \BracketRank achieves \textbf{26.56 nDCG@10}, significantly outperforming state-of-the-art baselines including RankGPT-4 (17.0) and Rank-R1-14B (20.5). On TREC datasets, BracketRank achieves 77.90 nDCG@5 on DL 19 and 75.85 nDCG@5 on DL 20, exceeding all baselines, establishing that explicit reasoning within competitive elimination is a powerful paradigm for complex, multi-step retrieval tasks. https://github.com/DataScienceUIBK/BracketRank
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。