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

TL;DR

BARE is a novel framework for one-tower visual grounding that mitigates modality biases and enhances semantic reasoning, achieving state-of-the-art results efficiently across multiple benchmarks.

Contribution

It introduces three modules to preserve modality-specific features, correct visual biases, and enhance referential relationships, advancing the capabilities of one-tower visual grounding models.

Findings

Achieves state-of-the-art performance on five benchmarks.

Demonstrates superior computational efficiency.

Effectively mitigates modality biases and improves referential understanding.

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.