Mitigating Hallucination in Visual-Language Models via Re-Balancing Contrastive Decoding

TL;DR

This paper introduces Re-Balancing Contrastive Decoding, a novel method that reduces hallucinations in visual-language models by recalibrating attention between visual and textual information, improving factual accuracy.

Contribution

The paper proposes a dual-branch attention re-balancing technique that mitigates hallucinations in VLMs without sacrificing their overall performance.

Findings

RBD outperforms existing methods on CHAIR and POPE metrics.

RBD effectively reduces hallucinations in VLMs.

The method maintains the models' general capabilities.

Abstract

Although Visual-Language Models (VLMs) have shown impressive capabilities in tasks like visual question answering and image captioning, they still struggle with hallucinations. Analysis of attention distribution in these models shows that VLMs tend to processing textual tokens rather than visual tokens. This imbalance of attention distribution causes VLMs to favor textual knowledge in the case of multimodal knowledge conflicts, resulting in differences from the image information. In this paper, we propose Re-Balancing Contrastive Decoding (RBD) method, which employs textual and visual branches to recalibrate attention distribution in VLMs. Specifically, the textual branch injects image noise to stimulate the model's dependency on text, thereby reducing textual bias. Concurrently, the visual branch focuses on the selection of significant tokens, refining the attention mechanism to highlight the primary subject. This dual-branch strategy enables the RBD method to diminish textual bias while enhancing visual information. Experimental results demonstrate that our method, RBD, outperforms the existing methods by the CHAIR and POPE metrics, mitigate hallucinations without reducing the model's general capabilities.