Multi-Stage Verification-Centric Framework for Mitigating Hallucination in Multi-Modal RAG

TL;DR

This paper introduces a multi-stage verification framework for multi-modal RAG systems that significantly reduces hallucinations and improves factual accuracy, demonstrated by achieving third place in a competitive benchmark.

Contribution

The paper presents a novel multi-stage framework with a query router, retrieval, summarization, and verification to mitigate hallucinations in multi-modal RAG models.

Findings

Achieved 3rd place in KDD Cup 2025 Meta CRAG-MM challenge.

Demonstrated reduced hallucinations and improved factual accuracy.

Validated effectiveness of multi-stage verification in complex multi-modal tasks.

Abstract

This paper presents the technical solution developed by team CRUISE for the KDD Cup 2025 Meta Comprehensive RAG Benchmark for Multi-modal, Multi-turn (CRAG-MM) challenge. The challenge aims to address a critical limitation of modern Vision Language Models (VLMs): their propensity to hallucinate, especially when faced with egocentric imagery, long-tail entities, and complex, multi-hop questions. This issue is particularly problematic in real-world applications where users pose fact-seeking queries that demand high factual accuracy across diverse modalities. To tackle this, we propose a robust, multi-stage framework that prioritizes factual accuracy and truthfulness over completeness. Our solution integrates a lightweight query router for efficiency, a query-aware retrieval and summarization pipeline, a dual-pathways generation and a post-hoc verification. This conservative strategy is designed to minimize hallucinations, which incur a severe penalty in the competition's scoring metric. Our approach achieved 3rd place in Task 1, demonstrating the effectiveness of prioritizing answer reliability in complex multi-modal RAG systems. Our implementation is available at https://github.com/Breezelled/KDD-Cup-2025-Meta-CRAG-MM .