通过定位并修正注意力头，提升CLIP模型在偏见数据上的公平性与性能。
Debiasing CLIP: Interpreting and Correcting Bias in Attention Heads
- 基于机制分析定位导致偏见的注意力头，再针对性移除。
- 在存在背景和性别偏见的数据集上，最差组准确率提升超50%。
- 可解释性强，适合关注AI公平性与模型可解释性的研究者。
多模态模型如CLIP因其在各类任务中的零样本表现而备受关注。然而研究发现，CLIP可能无意中学习到目标变量与混淆因素之间的虚假关联。为此，我们提出 extsc{Locate-Then-Correct}（LTC）框架，通过机制洞察识别视觉变换器中的虚假注意力头，并通过定向消融加以缓解。同时，LTC还能识别出对任务关键的显著注意力头，通过正交投影融合判别特征以提升分类性能。我们在具有内在背景和性别偏见的基准上评估了LTC，相比非训练后处理基线，最差组准确率提升超过50%。此外，我们可视化选定注意力头的表示，结果证实了该对比机制在识别虚假与显著注意力头方面的有效性。代码已公开于https://github.com/wj210/CLIP_LTC。
原文摘要 · Abstract (English)
Multimodal models like CLIP have gained significant attention due to their remarkable zero-shot performance across various tasks. However, studies have revealed that CLIP can inadvertently learn spurious associations between target variables and confounding factors. To address this, we introduce \textsc{Locate-Then-Correct} (LTC), a contrastive framework that identifies spurious attention heads in Vision Transformers via mechanistic insights and mitigates them through targeted ablation. Furthermore, LTC identifies salient, task-relevant attention heads, enabling the integration of discriminative features through orthogonal projection to improve classification performance. We evaluate LTC on benchmarks with inherent background and gender biases, achieving over a $>50\%$ gain in worst-group accuracy compared to non-training post-hoc baselines. Additionally, we visualize the representation of selected heads and find that the presented interpretation corroborates our contrastive mechanism for identifying both spurious and salient attention heads. Code available at https://github.com/wj210/CLIP_LTC.
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。