HELPD: Mitigating Hallucination of LVLMs by Hierarchical Feedback Learning with Vision-enhanced Penalty Decoding

TL;DR

The paper introduces HELPD, a hierarchical feedback learning framework with vision-enhanced penalty decoding, significantly reducing hallucinations in large vision-language models while improving output quality.

Contribution

It presents a novel hierarchical feedback mechanism and penalty decoding method that effectively mitigates multimodal hallucination in LVLMs with minimal training.

Findings

Reduces hallucination by over 15% across benchmarks.

Improves text generation quality in LVLMs.

Seamlessly integrates with existing LVLMs.

Abstract

Large Vision-Language Models (LVLMs) have shown remarkable performance on many visual-language tasks. However, these models still suffer from multimodal hallucination, which means the generation of objects or content that violates the images. Many existing work detects hallucination by directly judging whether an object exists in an image, overlooking the association between the object and semantics. To address this issue, we propose Hierarchical Feedback Learning with Vision-enhanced Penalty Decoding (HELPD). This framework incorporates hallucination feedback at both object and sentence semantic levels. Remarkably, even with a marginal degree of training, this approach can alleviate over 15% of hallucination. Simultaneously, HELPD penalizes the output logits according to the image attention window to avoid being overly affected by generated text. HELPD can be seamlessly integrated with any LVLMs. Our experiments demonstrate that the proposed framework yields favorable results across multiple hallucination benchmarks. It effectively mitigates hallucination for different LVLMs and concurrently improves their text generation quality.