2026/04/27 by Xihang Wang, Zihan Wang, Chengkai Huang +2 · 1 voice
Computer Science · #Ground #Key (lock) #Noise (video) #Pattern recognition (psychology) #Selection (genetic algorithm) #cs.CL #cs.IR #cs.IT
paper · pdf · doi:10.1145/3805712.3809947
arxiv published 2026/04/27 · arxiv updated 2026/04/30 · openalex publication_date 2026/07/15 · openalex created_date 2026/07/16 · openalex updated_date 2026/07/22
Multimodal Retrieval-Augmented Generation (MRAG) addresses key limitations of Multimodal Large Language Models (MLLMs), such as hallucination and outdated knowledge. However, current MRAG systems struggle to distinguish whether retrieved multimodal data truly supports the semantic core of an answer or merely provides superficial relevance. Existing metrics often rely on heuristic position-based confidence, which fails to capture the informational density of multimodal entities. To address this, we propose Multi-modal Evidence Grounding (MEG), a semantic-aware metric that quantifies the contribution of retrieved evidence. Unlike standard confidence measures, MEG utilizes Semantic Certainty Anchoring, which dynamically filters out high-frequency stopwords via Inverse Document Frequency (IDF) to focus strictly on information-bearing tokens. Building on MEG, we introduce MEG-RAG, a framework that trains a multimodal reranker to align retrieved evidence with the semantic anchors of the ground truth. By prioritizing high-value content based on semantic grounding rather than token probability distributions, MEG-RAG improves the accuracy and multimodal consistency of generated outputs. Extensive experiments on the ??2RAG benchmark show that MEG-RAG consistently outperforms strong baselines and demonstrates robust generalization across different teacher models. The data and code are available at here.