On Listwise Reranking for Corpus Feedback

2025-10-01

AI Summary

Existing document re-ranking methods face a fundamental trade-off: graph-aware approaches (e.g., GAR) rely on pre-constructed document similarity graphs, which are often unavailable and incur substantial memory overhead; graph-free methods, in contrast, require frequent LLM invocations, leading to high computational cost. This work proposes L2G, the first framework that implicitly induces document graph structure from list-level re-ranking logs—without explicit graph construction or additional LLM calls. Its core innovation lies in transforming historical re-ranking signals into lightweight, scalable graph representations and integrating them with LLM-derived semantic signals for graph-based reasoning. Evaluated on TREC-DL and BEIR subsets, L2G achieves retrieval performance comparable to oracle graph-based methods while significantly reducing both computational and storage overhead. Thus, L2G enables efficient, practical, and truly graph-aware re-ranking.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Natural Language Processing: Sentence-level Semantics, Textual Inference, etc.
• Knowledge Representation and Reasoning: Computational Complexity of Reasoning

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Reranker improves retrieval performance by capturing document interactions. At one extreme, graph-aware adaptive retrieval (GAR) represents an information-rich regime, requiring a pre-computed document similarity graph in reranking. However, as such graphs are often unavailable, or incur quadratic memory costs even when available, graph-free rerankers leverage large language model (LLM) calls to achieve competitive performance. We introduce L2G, a novel framework that implicitly induces document graphs from listwise reranker logs. By converting reranker signals into a graph structure, L2G enables scalable graph-based retrieval without the overhead of explicit graph computation. Results on the TREC-DL and BEIR subset show that L2G matches the effectiveness of oracle-based graph methods, while incurring zero additional LLM calls.

Problem

• Implicitly induces document graphs from reranker logs
• Enables scalable graph-based retrieval without explicit computation
• Matches oracle graph methods effectiveness with zero LLM calls

Innovation

• Implicitly induces document graphs from reranker logs
• Converts reranker signals into scalable graph structure
• Matches oracle graph methods without extra LLM calls