Less is More for RAG: Information Gain Pruning for Generator-Aligned Reranking and Evidence Selection

2026-01-24

AI Summary

This work addresses the weak or even negative correlation between traditional retrieval relevance metrics and answer quality under limited context budgets, which hinders effective evidence selection in retrieval-augmented generation (RAG) systems. To overcome this, the authors propose Information Gain Pruning (IGP), a novel approach that introduces generator-aligned utility signals to rerank and prune retrieved passages, eliminating redundant or harmful evidence. IGP requires no modification to existing context interfaces, ensuring straightforward deployment. Evaluated on five open-domain question answering benchmarks, IGP improves average F1 scores by 12%–20% while reducing input token consumption by 76%–79%, substantially enhancing the trade-off between generation quality and computational cost.

Technology Category

• Natural Language Processing: Discourse, Pragmatics & Argument Mining
• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval
• Knowledge Representation and Reasoning: Reasoning with Beliefs

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Retrieval-augmented generation (RAG) grounds large language models with external evidence, but under a limited context budget, the key challenge is deciding which retrieved passages should be injected. We show that retrieval relevance metrics (e.g., NDCG) correlate weakly with end-to-end QA quality and can even become negatively correlated under multi-passage injection, where redundancy and mild conflicts destabilize generation. We propose \textbf{Information Gain Pruning (IGP)}, a deployment-friendly reranking-and-pruning module that selects evidence using a generator-aligned utility signal and filters weak or harmful passages before truncation, without changing existing budget interfaces. Across five open-domain QA benchmarks and multiple retrievers and generators, IGP consistently improves the quality--cost trade-off. In a representative multi-evidence setting, IGP delivers about +12--20% relative improvement in average F1 while reducing final-stage input tokens by roughly 76--79% compared to retriever-only baselines.

Problem

• Retrieval-Augmented Generation
• Evidence Selection
• Context Budget
• Redundancy
• Generation Quality

Innovation

• Information Gain Pruning
• Retrieval-Augmented Generation
• Generator-Aligned Reranking
• Evidence Selection
• Context Budget Optimization