不需重训练，用向量修正分类模型偏见
No Training Wheels: Steering Vectors for Bias Correction at Inference Time
- 通过计算主流与少数群体激活均值差，生成纠偏向量
- 在推理时减去该向量，显著提升少数群体准确率
- 方法廉价高效，适用于各类分类模型
在群体代表性不均的数据集上训练的神经网络分类器常继承类别偏见并学习虚假关联。这些模型平均表现良好，但在少数群体上持续失败。例如，发色分类数据集可能过度代表女性金发，强化刻板印象。尽管已有算法和数据驱动方法缓解此类偏见，但通常需要重训练或大量算力。本文提出一种低成本、无需训练的方法，灵感来自大语言模型中的编辑向量。我们计算主流与少数群体激活均值之差，构建“偏见向量”，将其从模型残差流中减去。该方法有效降低分类偏见并提升最差群体准确率。我们探索了多种在类Transformer分类器中提取和应用该向量的策略，证明传统用于生成模型的引导向量也可在分类任务中生效。更广泛而言，本工作展示了一种极低成本、推理阶段、无需训练的分类模型偏见缓解方案。
原文摘要 · Abstract (English)
Neural network classifiers trained on datasets with uneven group representation often inherit class biases and learn spurious correlations. These models may perform well on average but consistently fail on atypical groups. For example, in hair color classification, datasets may over-represent females with blond hair, reinforcing stereotypes. Although various algorithmic and data-centric methods have been proposed to address such biases, they often require retraining or significant compute. In this work, we propose a cheap, training-free method inspired by steering vectors used to edit behaviors in large language models. We compute the difference in mean activations between majority and minority groups to define a "bias vector," which we subtract from the model's residual stream. This leads to reduced classification bias and improved worst-group accuracy. We explore multiple strategies for extracting and applying these vectors in transformer-like classifiers, showing that steering vectors, traditionally used in generative models, can also be effective in classification. More broadly, we showcase an extremely cheap, inference time, training free method to mitigate bias in classification models.
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。