Saved in:
Bibliographic Details
Main Authors: Hu, Xinmiao, Wang, Chun, An, Ruihe, Shao, ChenYu, Ye, Xiaojun, Zhou, Sheng, Li, Liangcheng
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2505.19474
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915305068953600
author Hu, Xinmiao
Wang, Chun
An, Ruihe
Shao, ChenYu
Ye, Xiaojun
Zhou, Sheng
Li, Liangcheng
author_facet Hu, Xinmiao
Wang, Chun
An, Ruihe
Shao, ChenYu
Ye, Xiaojun
Zhou, Sheng
Li, Liangcheng
contents Multimodal Large Language Models (MLLMs) have demonstrated strong performance in visual understanding tasks, yet they often suffer from object hallucinations--generating descriptions of objects that are inconsistent with or entirely absent from the input. This issue is closely related to dataset biases, where frequent co-occurrences of objects lead to entangled semantic representations across modalities. As a result, models may erroneously activate object representations that are commonly associated with the input but not actually present. To address this, we propose a causality-driven disentanglement framework that mitigates hallucinations through causal intervention. Our approach includes a Causal-Driven Projector in the visual pathway and a Causal Intervention Module integrated into the final transformer layer of the language model. These components work together to reduce spurious correlations caused by biased training data. Experimental results show that our method significantly reduces hallucinations while maintaining strong performance on multiple multimodal benchmarks. Visualization analyses further confirm improved separability of object representations. The code is available at: https://github.com/IgniSavium/Causal-LLaVA
format Preprint
id arxiv_https___arxiv_org_abs_2505_19474
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Causal-LLaVA: Causal Disentanglement for Mitigating Hallucination in Multimodal Large Language Models
Hu, Xinmiao
Wang, Chun
An, Ruihe
Shao, ChenYu
Ye, Xiaojun
Zhou, Sheng
Li, Liangcheng
Artificial Intelligence
Multimodal Large Language Models (MLLMs) have demonstrated strong performance in visual understanding tasks, yet they often suffer from object hallucinations--generating descriptions of objects that are inconsistent with or entirely absent from the input. This issue is closely related to dataset biases, where frequent co-occurrences of objects lead to entangled semantic representations across modalities. As a result, models may erroneously activate object representations that are commonly associated with the input but not actually present. To address this, we propose a causality-driven disentanglement framework that mitigates hallucinations through causal intervention. Our approach includes a Causal-Driven Projector in the visual pathway and a Causal Intervention Module integrated into the final transformer layer of the language model. These components work together to reduce spurious correlations caused by biased training data. Experimental results show that our method significantly reduces hallucinations while maintaining strong performance on multiple multimodal benchmarks. Visualization analyses further confirm improved separability of object representations. The code is available at: https://github.com/IgniSavium/Causal-LLaVA
title Causal-LLaVA: Causal Disentanglement for Mitigating Hallucination in Multimodal Large Language Models
topic Artificial Intelligence
url https://arxiv.org/abs/2505.19474