Bayesian Data Reweighting Improves Multimodal Retrieval for Knowledge-Based Visual Question Answering
Abstract
Multimodal retrievers are essential for knowledge-based visual question answering, where they retrieve external evidence for image-question pairs. However, existing contrastive training methods typically treat all unmatched query-document pairs as equally informative negatives, which is problematic because many unmatched documents may still be semantically relevant or partially useful. We propose Bayesian Data Reweighting, a probabilistic framework that models query-document importance as latent variables and adaptively infers posterior weights to downweight likely false negatives. With closed-form posterior updates under conjugate priors and stochastic EM optimization, our method consistently improves retrieval accuracy across three retrievers and seven knowledge-based VQA benchmarks.
Cite
@article{arxiv.2608.02907,
title = {Bayesian Data Reweighting Improves Multimodal Retrieval for Knowledge-Based Visual Question Answering},
author = {Jingchen Sun and Shaobo Han and Ruiyi Zhang and Naresh Kumar Devulapally and Ming Liu and Yitao Long and Vishnu Suresh Lokhande and Changyou Chen},
journal= {arXiv preprint arXiv:2608.02907},
year = {2026}
}