ActiShade: Activating Overshadowed Knowledge to Guide Multi-Hop Reasoning in Large Language Models

2026-01-12

AI Summary

This work addresses the challenge of multi-hop reasoning in large language models, where knowledge masking often leads to the retrieval of irrelevant information and error propagation. To mitigate this, the authors propose an iterative retrieval-augmented framework that introduces, for the first time, a mechanism to identify and activate masked keywords. By detecting missing critical information in the initial query, the method jointly retrieves evidence from multiple sources using both the original query and the identified keyword, then reconstructs the query based on retrieved results to guide subsequent reasoning steps. This process effectively supplements missing knowledge while suppressing noise. Experimental results demonstrate that the proposed approach significantly outperforms existing methods across multiple datasets and large language models, substantially improving both accuracy and robustness in multi-hop reasoning tasks.

Technology Category

• Natural Language Processing: Question Answering
• Knowledge Representation and Reasoning: Knowledge Acquisition
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

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

Abstract

In multi-hop reasoning, multi-round retrieval-augmented generation (RAG) methods typically rely on LLM-generated content as the retrieval query. However, these approaches are inherently vulnerable to knowledge overshadowing - a phenomenon where critical information is overshadowed during generation. As a result, the LLM-generated content may be incomplete or inaccurate, leading to irrelevant retrieval and causing error accumulation during the iteration process. To address this challenge, we propose ActiShade, which detects and activates overshadowed knowledge to guide large language models (LLMs) in multi-hop reasoning. Specifically, ActiShade iteratively detects the overshadowed keyphrase in the given query, retrieves documents relevant to both the query and the overshadowed keyphrase, and generates a new query based on the retrieved documents to guide the next-round iteration. By supplementing the overshadowed knowledge during the formulation of next-round queries while minimizing the introduction of irrelevant noise, ActiShade reduces the error accumulation caused by knowledge overshadowing. Extensive experiments show that ActiShade outperforms existing methods across multiple datasets and LLMs.

Problem

• multi-hop reasoning
• retrieval-augmented generation
• knowledge overshadowing
• error accumulation
• large language models

Innovation

• knowledge overshadowing
• multi-hop reasoning
• retrieval-augmented generation
• query refinement
• large language models