超越表面遗忘：多模态大语言模型中Hallucinations的锐度感知鲁棒擦除
Beyond Superficial Unlearning: Sharpness-Aware Robust Erasure of Hallucinations in Multimodal LLMs
AI总结：
本文提出SARE方法，通过目标导向的min-max优化和Targeted-SAM机制，解决多模态大语言模型中 hallucinations 的鲁棒擦除问题，提升模型稳定性与擦除效果。
摘要：
多模态大语言模型虽然强大，但容易产生hallucinations，即不存在的实体，影响可靠性。尽管最近的遗忘方法试图缓解这一问题，我们发现了一个关键缺陷：结构脆弱性。我们实证显示，标准擦除仅能表面抑制，使模型陷入尖锐极小值，轻度重新学习后hallucinations会灾难性复苏。为确保几何稳定性，我们提出SARE，将遗忘视为目标min-max优化问题，并使用Targeted-SAM机制显式平坦hallucinated概念周围的损失景观。通过在模拟最坏情况参数扰动下抑制hallucinations，我们的框架确保了鲁棒去除的稳定性。大量实验表明，SARE在擦除效果上显著优于基线，同时保持一般生成质量。关键的是，它在重新学习和参数更新中维持持久的hallucination抑制，验证了几何稳定性的有效性。
英文摘要：
Multimodal LLMs are powerful but prone to object hallucinations, which describe non-existent entities and harm reliability. While recent unlearning methods attempt to mitigate this, we identify a critical flaw: structural fragility. We empirically demonstrate that standard erasure achieves only superficial suppression, trapping the model in sharp minima where hallucinations catastrophically resurge after lightweight relearning. To ensure geometric stability, we propose SARE, which casts unlearning as a targeted min-max optimization problem and uses a Targeted-SAM mechanism to explicitly flatten the loss landscape around hallucinated concepts. By suppressing hallucinations under simulated worst-case parameter perturbations, our framework ensures robust removal stable against weight shifts. Extensive experiments demonstrate that SARE significantly outperforms baselines in erasure efficacy while preserving general generation quality. Crucially, it maintains persistent hallucination suppression against relearning and parameter updates, validating the effectiveness of geometric stabilization.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
