BARE: Towards Bias-Aware and Reasoning-Enhanced One-Tower Visual Grounding

2026-01-04

arXiv.org

AI Summary

This work addresses the limitations of existing single-tower visual grounding methods, which suffer from modality bias due to excessive entanglement of multimodal representations and insufficient semantic reasoning capabilities, leading to inaccurate interpretation of referring expressions. To overcome these challenges, the authors propose the BARE framework, which integrates three synergistic modules—linguistic salience modulation, visual bias correction, and referential relation enhancement—to strengthen referential semantic understanding while preserving modality-specific characteristics. Built upon a single-tower multimodal architecture, BARE achieves state-of-the-art performance across five benchmark datasets, demonstrating not only superior accuracy and robustness in visual grounding but also improved computational efficiency.

Technology Category

• Natural Language Processing: Language Grounding & Multi-modal NLP
• Computer Vision: Multi-modal Vision
• Machine Learning: Multimodal Learning

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Visual Grounding (VG), which aims to locate a specific region referred to by expressions, is a fundamental yet challenging task in the multimodal understanding fields. While recent grounding transfer works have advanced the field through one-tower architectures, they still suffer from two primary limitations: (1) over-entangled multimodal representations that exacerbate deceptive modality biases, and (2) insufficient semantic reasoning that hinders the comprehension of referential cues. In this paper, we propose BARE, a bias-aware and reasoning-enhanced framework for one-tower visual grounding. BARE introduces a mechanism that preserves modality-specific features and constructs referential semantics through three novel modules: (i) language salience modulator, (ii) visual bias correction and (iii) referential relationship enhancement, which jointly mitigate multimodal distractions and enhance referential comprehension. Extensive experimental results on five benchmarks demonstrate that BARE not only achieves state-of-the-art performance but also delivers superior computational efficiency compared to existing approaches. The code is publicly accessible at https://github.com/Marloweeee/BARE.

Problem

• visual grounding
• modality bias
• semantic reasoning
• one-tower architecture
• referential comprehension

Innovation

• bias-aware
• reasoning-enhanced
• one-tower architecture
• visual grounding
• modality-specific features