模型删数据后看似可靠，实则更依赖错误线索做决策。
Calibration vs Decision Making: Revisiting the Reliability Paradox in Unlearned Language Models
- 用删除数据后模型的校准度与归因分析对比，发现校准好不代表决策可靠。
- 微调模型校准误差仅0.04，但删数据后仍保持低误差却更依赖无关关键词。
- 揭示了机器删忆中的可靠性悖论，适合研究模型可信性的人看。
机器未学习旨在移除特定训练数据对模型的影响，同时保持剩余数据上的可靠预测与不确定性估计，因此可靠性评估至关重要。校准常被用作语言模型可靠性的代理指标，但低校准误差并不意味着决策规则可靠，因为模型可能依赖虚假相关性。我们通过TOFU基准上的多选问答任务，采用校准指标（ECE、MCE、Brier）衡量概率可靠性，并使用积分梯度和局部互信息进行归因分析检测捷径。结果表明，微调模型校准误差仅为ECE ~ 0.04，远低于预训练模型（ECE > 0.5），而删数据后模型虽准确率下降，校准误差仍维持低位，但归因分析显示其更依赖相关性特征。这说明删除数据后，良好校准可与基于捷径的决策共存，将可靠性悖论扩展至机器未学习场景。
原文摘要 · Abstract (English)
Machine unlearning aims to remove the influence of specific training data from a model while preserving reliable behavior on the remaining data, making reliable prediction and uncertainty estimation essential for evaluation. Calibration is commonly used as a proxy for reliability in language models, but low calibration error does not necessarily imply reliable decision rules, as models may rely on spurious correlations while remaining well calibrated. We investigate this gap in generative language models using the multiple-choice question-answering evaluation protocol on the TOFU benchmark, measuring probabilistic reliability with calibration metrics (ECE, MCE, Brier) and decision-rule reliability via attribution-based shortcut detection with Integrated Gradients and Local Mutual Information. We find that fine-tuned models achieve low calibration error (ECE ~ 0.04) compared to pretrained models (ECE > 0.5), and models after unlearning retain similarly low calibration despite reduced accuracy on the forget split, while attribution analysis shows increased reliance on correlation-based tokens. These results demonstrate that good calibration can coexist with shortcut-based decision rules after unlearning, extending the reliability paradox to the machine unlearning setting.
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。