用标注者分布训练模型，让预测不确定性更贴近人类判断。
Distributions In, Distributions Out: The Case for Soft-Label Training
- 用完整标注分布代替单一投票标签进行训练
- 准确率持平或更高，标注熵匹配度提升61%
- 适合处理人类主观分歧的任务，如语义推理、语气判断
监督分类器输出类别分布，但通常以多数投票合并多标注者意见，得到单一标签进行训练。在标注者存在分歧反映真实模糊性的任务中（如自然语言推理、礼貌性判断、视觉模糊分类），这种合并会丢弃信息，迫使模型对人类系统性不一致的输入表现出统一置信度。本文对比软标签训练（使用完整标注分布作为目标）与硬标签训练，在涵盖视觉和NLP的三个数据集（ChaosNLI、POPQUORN、CIFAR-10H）上进行实验。软标签训练在所有数据集上达到或超过硬标签准确率，平均降低32%的KL散度（p < 10^-4），且样本级预测熵与标注者熵的相关性提升61%——模型在人类不确定时也表现出相应不确定性。我们认为，当标注者合法分歧时，标注分布本身就是正确的学习目标，而非其噪声估计。
原文摘要 · Abstract (English)
Supervised classifiers output a distribution over classes but are typically trained against a single label obtained by collapsing multiple annotators into a majority vote. On tasks where annotator disagreement reflects genuine ambiguity -- natural language inference, politeness, visually ambiguous categorization -- this collapse discards information and forces models to express uniform confidence on inputs where humans systematically disagree. We compare soft-label training, which uses the full annotation distribution as the target, against hard-label training across three datasets spanning vision and NLP (ChaosNLI, POPQUORN, CIFAR-10H). Soft-label training matches or exceeds hard-label accuracy on every dataset, reduces KL divergence to the annotator distribution by 32% on average (p < 10^-4), and produces predictions whose per-sample entropy correlates 61% more strongly with annotator entropy -- models trained on distributions are uncertain precisely where humans are. We argue these benefits follow from a basic observation: when annotators legitimately disagree, the annotation distribution is the correct learning target, not a noisy estimate of it.
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。