Think Thrice Before Reranking: Multi-perspective Evidence and Reasoning Integration for Text Reranking
MERIT-Rank integrates multiple LLM reasoning trajectories for text reranking; its 4B model outperforms most 7B and 32B rerankers on BRIGHT.
MERIT-Rank formulates a Multi-Trajectory Reasoning Space that evaluates query-document relevance from multiple perspectives and consolidates reasoning paths with a joint reranker. A progressive training framework, Progressive Rank Policy Optimization (PRPO), stabilizes reasoning trajectories while improving ranking quality through staged objectives. Experiments on reasoning-intensive and traditional retrieval benchmarks show consistent gains over competitive baselines, with the 4B model outperforming most 7B and even 32B rerankers on BRIGHT.
- Multi-trajectory reasoning space evaluates relevance from multiple perspectives
- PRPO progressive training stabilizes reasoning with staged objectives
- 4B model outperforms most 7B and 32B rerankers on BRIGHT
- Gains consistent on reasoning-intensive and traditional retrieval benchmarks
Full article154 words · extracted from arxiv.org · click to collapse
Reasoning-based reranking with Large Language Models (LLMs) has shown promising improvements in text ranking. However, current methods predominantly rely on a single reasoning trajectory, resulting in rankings that are susceptible to reasoning errors and inherently constrained in modeling the multifaceted signals underlying document relevance. To resolve this dilemma, we propose MERIT-Rank(Multi-perspective Evidence and Reasoning Integration for Text Reranking), a framework that models complementary reasoning trajectories to improve reranking robustness. MERIT-Rank formulates a Multi-Trajectory Reasoning Space (MTRS) that evaluates query-document relevance from multiple perspectives and introduces a joint reranker that consolidates these reasoning paths into a unified ranking decision. We further develop Progressive Rank Policy Optimization (PRPO), a progressive training framework that stabilizes reasoning trajectories while continually improving ranking quality through staged optimization objectives. Experiments on both reasoning-intensive and traditional retrieval benchmarks show that MERIT-Rank consistently achieves superior performance over competitive baselines. The 4B model notably outperforms most 7B and even 32B rerankers on BRIGHT.
Text extracted automatically; images, tables and formatting may be missing. Original: https://arxiv.org/abs/2609.20131