Poison as Cure: Visual Noise for Mitigating Object Hallucinations in LVMs
This addresses reliability issues in LVMs for applications requiring accurate visual interpretation, though it is an incremental improvement as it modifies inputs rather than the model architecture.
The paper tackles object hallucination in large vision-language models (LVMs) by proposing a visual adversarial perturbation (VAP) method that applies optimized visual noise to reduce factually inaccurate outputs, achieving consistent reductions across 8 state-of-the-art LVMs.
Large vision-language models (LVMs) extend large language models (LLMs) with visual perception capabilities, enabling them to process and interpret visual information. A major challenge compromising their reliability is object hallucination that LVMs may generate plausible but factually inaccurate information. We propose a novel visual adversarial perturbation (VAP) method to mitigate this hallucination issue. VAP alleviates LVM hallucination by applying strategically optimized visual noise without altering the base model. Our approach formulates hallucination suppression as an optimization problem, leveraging adversarial strategies to generate beneficial visual perturbations that enhance the model's factual grounding and reduce parametric knowledge bias. Extensive experimental results demonstrate that our method consistently reduces object hallucinations across 8 state-of-the-art LVMs, validating its efficacy across diverse evaluations.