发现视觉指令微调模型虽被脏数据损坏，但可修复且自带辨别能力。
Corrupted but Not Broken: Understanding and Mitigating the Negative Impacts of Corrupted Data in Visual Instruction Tuning
- 通过关闭少量参数恢复模型性能，实现反脆弱训练。
- 脏数据下模型性能下降，但关闭部分参数后几乎完全恢复。
- 适合处理低质量多模态数据的场景，如医疗、工业图像理解。
视觉指令微调（VIT）旨在提升多模态大语言模型（MLLMs）性能，但其效果常因包含幻觉内容、错误回答和低质量OCR的脏数据而受损。以往方法依赖高质量数据收集或规则过滤，成本高且覆盖有限。本文系统研究脏数据对MLLMs的影响，发现尽管脏数据降低模型表现，但这种负面影响在很大程度上是可逆的，即模型具备‘被污染但未崩溃’的特性。具体而言，仅需禁用一小部分参数，即可近乎完全恢复性能；此外，被污染的MLLMs能自主区分干净与脏样本，实现无需外部干预的数据清洗。基于此，我们提出一种抗脏数据训练范式，显著优于现有策略。
原文摘要 · Abstract (English)
Visual Instruction Tuning (VIT) aims to enhance Multimodal Large Language Models (MLLMs), yet its effectiveness is often compromised by corrupted datasets with issues such as hallucinated content, incorrect responses, and poor OCR quality. Previous approaches to address these challenges have focused on refining datasets through high-quality data collection or rule-based filtering that can be costly or limited in scope. In this paper, we conduct a systematic investigation into the impact of corrupted data on MLLMs and discover that, although corrupted data degrade model performance, such adverse effects are largely reversible, and MLLMs are {\bf corrupted but not broken}. Specifically, we find that disabling a small subset of parameters can almost fully restore performance. Moreover, corrupted MLLMs inherently possess the capability to differentiate between clean and corrupted samples, facilitating dataset cleaning without external intervention. Building on these insights, we introduce a corruption-robust training paradigm that significantly surpasses existing strategies for mitigating the effects of corrupted data.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。