REAL: Resolving Knowledge Conflicts in Knowledge-Intensive Visual Question Answering via Reasoning-Pivot Alignment

TL;DR

The paper introduces the REAL framework for resolving knowledge conflicts in knowledge-intensive visual question answering by leveraging reasoning-pivots, which are external evidence-based units in the reasoning chain, leading to improved accuracy.

Contribution

The paper proposes the novel concept of reasoning-pivots and a framework that uses them for conflict detection and mitigation in KI-VQA, supported by a new dataset and decoding strategies.

Findings

Achieves state-of-the-art performance on KI-VQA benchmarks.

Significantly improves conflict discrimination accuracy.

Demonstrates effectiveness of pivot-driven conflict resolution.

Abstract

Knowledge-intensive Visual Question Answering (KI-VQA) frequently suffers from severe knowledge conflicts caused by the inherent limitations of open-domain retrieval. However, existing paradigms face critical limitations due to the lack of generalizable conflict detection and intra-model constraint mechanisms to handle conflicting evidence. To address these challenges, we propose the REAL (Reasoning-Pivot Alignment) framework centered on the novel concept of the Reasoning-Pivot. Distinct from reasoning steps that prioritize internal self-derivation, a reasoning-pivot serves as an atomic unit (node or edge) in the reasoning chain that emphasizes knowledge linkage, and it typically relies on external evidence to complete the reasoning. Supported by our constructed REAL-VQA dataset, our approach integrates Reasoning-Pivot Aware SFT (RPA-SFT) to train a generalizable discriminator by aligning conflicts with pivot extraction, and employs Reasoning-Pivot Guided Decoding (RPGD), an intra-model decoding strategy that leverages these pivots for targeted conflict mitigation. Extensive experiments across diverse benchmarks demonstrate that REAL significantly enhances discrimination accuracy and achieves state-of-the-art performance, validating the effectiveness of our pivot-driven resolution paradigm.