Mitigating Object Hallucination in Large Vision-Language Models via Image-Grounded Guidance

2024/02/13 by Linxi Zhao, Zhao, Linxi, Yihe Deng +5 · 15 citations
Computer Science · Medicine · Neuroscience · #Adversarial Robustness in Machine Learning #Artificial Intelligence (cs.AI) #Brain Tumor Detection and Classification #Computation and Language (cs.CL) #Computer Vision and Pattern Recognition (cs.CV) #Epilepsy research and treatment #FOS: Computer and information sciences #Machine Learning (cs.LG)

paper · pdf · doi:10.48550/arxiv.2402.08680

openalex publication_date 2024/02/13 · openalex created_date 2024/02/15 · openalex updated_date 2026/07/28

Abstract

The advancement of Large Vision-Language Models (LVLMs) has increasingly highlighted the critical issue of their tendency to hallucinate non-existing objects in the images. To address this issue, previous works focused on using specially curated datasets or powerful LLMs to rectify the outputs of LVLMs. However, these approaches require either costly training or fine-tuning, or API access to proprietary LLMs for post-generation correction. In response to these limitations, we propose Mitigating hallucinAtion via image-gRounded guIdaNcE (MARINE), a framework that is both training-free and API-free. MARINE effectively and efficiently reduces object hallucinations during inference by introducing image-grounded guidance to LVLMs. This is achieved by leveraging open-source vision models to extract object-level information, thereby enhancing the precision of LVLM-generated content. Our framework's flexibility further allows for the integration of multiple vision models, enabling more reliable and robust object-level guidance. Through comprehensive evaluations across 5 popular LVLMs with diverse evaluation metrics and benchmarks, we demonstrate the effectiveness of MARINE, which even outperforms existing fine-tuning-based methods. Remarkably, it reduces hallucinations consistently in GPT-4V-assisted evaluation while maintaining the detailedness of LVLMs' generations. We release our code at https://github.com/Linxi-ZHAO/MARINE.

Cited by

• EraseLoRA: MLLM-Driven Foreground Exclusion and Background Subtype Aggregation for Dataset-Free Object Removal
• SEASON: Mitigating Temporal Hallucination in Video Large Language Models via Self-Diagnostic Contrastive Decoding
• Tell Model Where to Look: Mitigating Hallucinations in MLLMs by Vision-Guided Attention
• Adaptive Residual-Update Steering for Low-Overhead Hallucination Mitigation in Large Vision Language Models
• "It's trained by non-disabled people": Evaluating How Image Quality Affects Product Captioning with VLMs
• SHIELD: Suppressing Hallucinations In LVLM Encoders via Bias and Vulnerability Defense
• HARMONY: Hidden Activation Representations and Model Output-Aware Uncertainty Estimation for Vision-Language Models
• Token-Level Inference-Time Alignment for Vision-Language Models
• Reallocating Attention Across Layers to Reduce Multimodal Hallucination
• Mitigating Diffusion Model Hallucinations with Dynamic Guidance
• CCD: Mitigating Hallucinations in Radiology MLLMs via Clinical Contrastive Decoding
• Losing the Plot: How VLM responses degrade on imperfect charts
• Mitigating Multimodal Hallucinations via Gradient-based Self-Reflection
• OmniDPO: A Preference Optimization Framework to Address Omni-Modal Hallucination
• Controlling Multimodal LLMs via Reward-guided Decoding

Related

• Analyzing and Mitigating Object Hallucination in Large Vision-Language Models
• Seeing is Believing: Mitigating Hallucination in Large Vision-Language Models via CLIP-Guided Decoding
• Does Object Grounding Really Reduce Hallucination of Large Vision-Language Models?
• ICT: Image-Object Cross-Level Trusted Intervention for Mitigating Object Hallucination in Large Vision-Language Models
• Mitigating Hallucination in Visual Language Models with Visual Supervision