Seeing through the Conflict: Transparent Knowledge Conflict Handling in Retrieval-Augmented Generation

2026-01-11

arXiv.org

AI Summary

This work addresses the tendency of large language models in retrieval-augmented generation (RAG) to conflate internal parametric knowledge with external evidence, leading to hallucinations or neglect of critical context. To mitigate this, the authors propose the TCR framework, which employs dual contrastive encoders to disentangle semantic matching from factual consistency, incorporates a self-answering assessment to determine whether the model can respond reliably without retrieved evidence, and integrates a lightweight soft prompt weighted by signal-to-noise ratio (SNR) to inject three complementary signals into the generator for transparent and controllable conflict resolution. With only a 0.3% parameter increase, TCR achieves significant performance gains across seven benchmarks: improving conflict detection F1 by 5–18 points, increasing knowledge gap recovery by 21.4%, reducing reliance on misleading context by 29.3%, and demonstrating high alignment between its decision signals and human judgments.

Technology Category

• Natural Language Processing: Language Grounding & Multi-modal NLP
• Machine Learning: Calibration & Uncertainty Quantification
• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval

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

Large language models (LLMs) equipped with retrieval--the Retrieval-Augmented Generation (RAG) paradigm--should combine their parametric knowledge with external evidence, yet in practice they often hallucinate, over-trust noisy snippets, or ignore vital context. We introduce TCR (Transparent Conflict Resolution), a plug-and-play framework that makes this decision process observable and controllable. TCR (i) disentangles semantic match and factual consistency via dual contrastive encoders, (ii) estimates self-answerability to gauge confidence in internal memory, and (iii) feeds the three scalar signals to the generator through a lightweight soft-prompt with SNR-based weighting. Across seven benchmarks TCR improves conflict detection (+5-18 F1), raises knowledge-gap recovery by +21.4 pp and cuts misleading-context overrides by -29.3 pp, while adding only 0.3% parameters. The signals align with human judgements and expose temporal decision patterns.

Problem

• Retrieval-Augmented Generation
• knowledge conflict
• hallucination
• factual consistency
• transparent decision-making

Innovation

• Retrieval-Augmented Generation
• Knowledge Conflict Resolution
• Contrastive Encoding
• Self-Answerability
• Soft-Prompting