arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2603.15656cs.LGcs.AIcs.CV
收藏

基于归因的不可靠神经网络行为模型修正

Attribution-Guided Model Rectification of Unreliable Neural Network Behaviors

The University of Melbourne(墨尔本大学);The University of Western Australia(西澳大学)

Peiyu Yang, Naveed Akhtar, Jiantong Jiang, Ajmal Mian

AI总结:

本文提出基于归因的模型修正框架,通过秩一模型编辑有效定位并修正模型不可靠行为,减少对清洁样本的依赖,提升模型鲁棒性。

摘要:

神经网络模型性能下降是由于其在受污染样本非鲁棒特征上的不可靠行为所致。由于其不透明性,修正模型以解决此问题通常需要繁琐的数据清洗和模型重训练,导致巨大的计算和人工开销。本文利用秩一模型编辑建立一个基于归因的模型修正框架,有效定位并修正模型不可靠行为。我们首先区分我们的修正设置与现有模型编辑,得出一种在保持模型性能的同时减少对清洁样本预算依赖的公式。我们进一步揭示了模型修正中由于层间编辑性异质性而产生的瓶颈。为针对主要的误行源,我们引入了基于归因的层定位方法,量化各层的编辑性并识别最负责不可靠性的层。大量实验显示,本文方法在纠正神经毒药、虚假相关性和特征泄漏中表现出色。我们的方法通过仅使用单个清洁样本即可达成编辑目标,使其在实践中具有吸引力。

英文摘要:

The performance of neural network models deteriorates due to their unreliable behavior on non-robust features of corrupted samples. Owing to their opaque nature, rectifying models to address this problem often necessitates arduous data cleaning and model retraining, resulting in huge computational and manual overhead. In this work, we leverage rank-one model editing to establish an attribution-guided model rectification framework that effectively locates and corrects model unreliable behaviors. We first distinguish our rectification setting from existing model editing, yielding a formulation that corrects unreliable behavior while preserving model performance and reducing reliance on large budgets of cleansed samples. We further reveal a bottleneck of model rectifying arising from heterogeneous editability across layers. To target the primary source of misbehavior, we introduce an attribution-guided layer localization method that quantifies layer-wise editability and identifies the layer most responsible for unreliabilities. Extensive experiments demonstrate the effectiveness of our method in correcting unreliabilities observed for neural Trojans, spurious correlations and feature leakage. Our method shows remarkable performance by achieving its editing objective with as few as a single cleansed sample, which makes it appealing for practice.

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。

↑