研究人员开发了一种新的黑盒辅助回归方法，旨在提高基础模型在下游任务中使用有限数据时的可靠性。该方法称为安全残差估计器（Safe Residual Estimator），它学习对现有黑盒预测器进行修正，并使用验证数据，如果修正支持不佳，则恢复到原始预测器，从而避免负迁移。在合成数据和真实世界数据集（如 CIFAR-100 和 AG News）上使用 CLIP 和 Qwen3-8B 等模型进行的实验，证明了这种残差修正权衡的有效性。 AI
影响 增强了在下游任务中使用有限数据的基础模型的可靠性和安全性。
排序理由 该集群包含一篇详细介绍新研究方法和实验结果的学术论文。
AI 生成摘要 · Google Gemini · 来自 2 个来源。 我们如何撰写摘要 →
arXiv:2606.25743v1 Announce Type: new Abstract: Foundation models are often used as fixed black-box predictors for downstream tasks with limited labeled data, but their predictions may be biased and unsafe to trust blindly. We study this setting through black-box assisted nonpara…
Foundation models are often used as fixed black-box predictors for downstream tasks with limited labeled data, but their predictions may be biased and unsafe to trust blindly. We study this setting through black-box assisted nonparametric regression: a learner observes labeled sa…