Detecting and Mitigating Hallucination in Large Vision Language Models via Fine-Grained AI Feedback

TL;DR

This paper introduces a fine-grained AI feedback approach to detect and reduce hallucinations in large vision language models, improving alignment with given contexts through a novel detect-then-rewrite pipeline and severity-aware optimization.

Contribution

It presents a new sentence-level hallucination detection model and a severity-aware preference optimization method for better hallucination mitigation in LVLMs.

Findings

Effective hallucination detection at sentence level

Improved hallucination mitigation with severity-aware optimization

Enhanced model alignment with context in experiments

Abstract

The rapidly developing Large Vision Language Models (LVLMs) have shown notable capabilities on a range of multi-modal tasks, but still face the hallucination phenomena where the generated texts do not align with the given contexts, significantly restricting the usages of LVLMs. Most previous work detects and mitigates hallucination at the coarse-grained level or requires expensive annotation (e.g., labeling by proprietary models or human experts). To address these issues, we propose detecting and mitigating hallucinations in LVLMs via fine-grained AI feedback. The basic idea is that we generate a small-size sentence-level hallucination annotation dataset by proprietary models, whereby we train a hallucination detection model which can perform sentence-level hallucination detection, covering primary hallucination types (i.e., object, attribute, and relationship). Then, we propose a detect-then-rewrite pipeline to automatically construct preference dataset for training hallucination mitigating model. Furthermore, we propose differentiating the severity of hallucinations, and introducing a Hallucination Severity-Aware Direct Preference Optimization (HSA-DPO) for mitigating hallucination in LVLMs by incorporating the severity of hallucinations into preference learning. Extensive experiments demonstrate the effectiveness of our method.