Mitigating Hallucination in Visual Language Models with Visual Supervision

TL;DR

This paper proposes methods to reduce hallucination in vision-language models by incorporating detailed visual annotations, auxiliary supervision, and a new benchmark for evaluation, leading to more accurate multi-modal responses.

Contribution

It introduces detailed relationship annotations, integrates SAM and mask prediction as supervision, and develops RAH-Bench for comprehensive hallucination evaluation.

Findings

8.4% improvement over LLaVA in hallucination mitigation

Enhanced accuracy in detailed image understanding

Widespread performance gains across multiple models

Abstract

Large vision-language models (LVLMs) suffer from hallucination a lot, generating responses that apparently contradict to the image content occasionally. The key problem lies in its weak ability to comprehend detailed content in a multi-modal context, which can be mainly attributed to two factors in training data and loss function. The vision instruction dataset primarily focuses on global description, and the auto-regressive loss function favors text modeling rather than image understanding. In this paper, we bring more detailed vision annotations and more discriminative vision models to facilitate the training of LVLMs, so that they can generate more precise responses without encounter hallucination. On one hand, we generate image-text pairs with detailed relationship annotations in panoptic scene graph dataset (PSG). These conversations pay more attention on detailed facts in the image, encouraging the model to answer questions based on multi-modal contexts. On the other hand, we integrate SAM and mask prediction loss as auxiliary supervision, forcing the LVLMs to have the capacity to identify context-related objects, so that they can generate more accurate responses, mitigating hallucination. Moreover, to provide a deeper evaluation on the hallucination in LVLMs, we propose a new benchmark, RAH-Bench. It divides vision hallucination into three different types that contradicts the image with wrong categories, attributes or relations, and introduces False Positive Rate as detailed sub-metric for each type. In this benchmark, our approach demonstrates an +8.4% enhancement compared to original LLaVA and achieves widespread performance improvements across other models.