Beyond Visual Similarity: Entity-Aligned Retrieval for Knowledge-Based Visual Question Answering

Hugging Face Daily Papers Papers

Summary

KBMR is an MLLM-based retriever for knowledge-based visual question answering that uses semantic alignment to improve retrieval and VQA accuracy, showing significant gains over CLIP baselines.

Knowledge-Based Visual Question Answering (KB-VQA) relies on retrieving external information to answer queries involving long-tail entities. However, existing retrieval pipelines predominantly employ CLIP-style dual encoders, which prioritize surface-level visual similarity over entity-level semantic alignment. This paradigm often fails when semantically identical concepts exhibit large visual variations or when distinct entities appear visually similar. To address this, we propose KBMR, the first MLLM-based embedding retriever tailored for KB-VQA. Leveraging the robust autoregressive capabilities of MLLMs, KBMR maps images into a semantic space that better preserves concept identity. To tackle the challenge of noisy supervision in Wikipedia-scale retrieval, we introduce an MLLM-based semantic discriminator that generates continuous entity-consistency weights. These weights guide a novel continuous semantic distillation objective, enabling effective hard negative sampling and soft supervision beyond rigid binary labels. Extensive experiments demonstrate that KBMR significantly outperforms CLIP baselines, yielding up to a 14.7% improvement in retrieval Recall@1 and a 9.4% gain in end-to-end VQA accuracy. Code is available at https://github.com/realHarryX/KBMR.
Original Article
View Cached Full Text

Cached at: 09/03/26, 11:52 AM

Paper page - Beyond Visual Similarity: Entity-Aligned Retrieval for Knowledge-Based Visual Question Answering

Source: https://huggingface.co/papers/2608.21450

Abstract

KBMR uses a multimodal language model to embed images by semantic identity rather than surface appearance, improving retrieval and visual question answering via continuous distillation and hard negative sampling.

Knowledge-Based Visual Question Answering (KB-VQA) relies on retrieving external information to answer queries involving long-tail entities. However, existing retrieval pipelines predominantly employCLIP-style dual encoders, which prioritize surface-level visual similarity over entity-level semantic alignment. This paradigm often fails when semantically identical concepts exhibit large visual variations or when distinct entities appear visually similar. To address this, we propose KBMR, the firstMLLM-based embedding retrievertailored forKB-VQA. Leveraging the robustautoregressive capabilitiesof MLLMs, KBMR maps images into a semantic space that better preserves concept identity. To tackle the challenge of noisy supervision in Wikipedia-scale retrieval, we introduce an MLLM-basedsemantic discriminatorthat generatescontinuous entity-consistency weights. These weights guide a novelcontinuous semantic distillationobjective, enabling effectivehard negative samplingand soft supervision beyond rigid binary labels. Extensive experiments demonstrate that KBMR significantly outperforms CLIP baselines, yielding up to a 14.7% improvement in retrieval Recall@1 and a 9.4% gain in end-to-end VQA accuracy. Code is available at https://github.com/realHarryX/KBMR.

View arXiv pageView PDFAdd to collection

Get this paper in your agent:

hf papers read 2608\.21450

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2608.21450 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2608.21450 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2608.21450 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles