Hallucination-aware intermediate representation edit in large vision-language models
This addresses the problem of unreliable outputs in vision-language models for users in applications like multimodal reasoning, offering an efficient solution compared to resource-intensive retraining or dual-inference methods.
The paper tackles hallucination issues in large vision-language models, where outputs contradict visual facts, by proposing a framework for dynamically detecting and editing hallucination representations with minimal computational cost, achieving state-of-the-art performance on benchmarks.
Large Vision-Language Models have demonstrated exceptional performance in multimodal reasoning and complex scene understanding. However, these models still face significant hallucination issues, where outputs contradict visual facts. Recent research on hallucination mitigation has focused on retraining methods and Contrastive Decoding (CD) methods. While both methods perform well, retraining methods require substantial training resources, and CD methods introduce dual inference overhead. These factors hinder their practical applicability. To address the above issue, we propose a framework for dynamically detecting hallucination representations and performing hallucination-eliminating edits on these representations. With minimal additional computational cost, we achieve state-of-the-art performance on existing benchmarks. Extensive experiments demonstrate the effectiveness of our approach, highlighting its efficient and robust hallucination elimination capability and its powerful controllability over hallucinations. Code is available at https://github.com/ASGO-MM/HIRE