Modeling Contextual Passage Utility for Multihop Question Answering

2025-12-06

AI Summary

In multi-hop question answering, retrieved passages often contain redundancy, degrading answer accuracy. Existing utility prediction methods treat passages in isolation, ignoring inter-passage dependencies and thus failing to model their complementary or connective roles within reasoning chains. This paper proposes a context-aware passage utility modeling approach: it explicitly encodes passage usage order and interaction patterns derived from reasoning trajectories, and employs a lightweight Transformer trained on synthetic data for fine-grained utility prediction. Crucially, it shifts utility modeling from isolated relevance assessment to context-dependent reasoning-chain modeling for retrieval reranking. Experiments demonstrate substantial improvements over relevance-based baselines across multiple multi-hop QA benchmarks, yielding higher answer accuracy and robustness while effectively suppressing redundant passage interference.

Technology Category

• Reasoning under Uncertainty: Decision/Utility Theory
• Knowledge Representation and Reasoning: Preferences
• Planning, Routing, and Scheduling: Model-Based Reasoning

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• 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

Multihop Question Answering (QA) requires systems to identify and synthesize information from multiple text passages. While most prior retrieval methods assist in identifying relevant passages for QA, further assessing the utility of the passages can help in removing redundant ones, which may otherwise add to noise and inaccuracies in the generated answers. Existing utility prediction approaches model passage utility independently, overlooking a critical aspect of multihop reasoning: the utility of a passage can be context-dependent, influenced by its relation to other passages - whether it provides complementary information or forms a crucial link in conjunction with others. In this paper, we propose a lightweight approach to model contextual passage utility, accounting for inter-passage dependencies. We fine-tune a small transformer-based model to predict passage utility scores for multihop QA. We leverage the reasoning traces from an advanced reasoning model to capture the order in which passages are used to answer a question and obtain synthetic training data. Through comprehensive experiments, we demonstrate that our utility-based scoring of retrieved passages leads to improved reranking and downstream QA performance compared to relevance-based reranking methods.

Problem

• Modeling contextual passage utility for multihop QA
• Assessing passage utility to reduce redundancy and noise
• Accounting for inter-passage dependencies in utility prediction

Innovation

• Model contextual passage utility with inter-passage dependencies
• Fine-tune small transformer for utility scores in QA
• Use reasoning traces to generate synthetic training data