الملخص
Deep neural networks, including large language models, have achieved remarkable performance across various tasks. However, they are prone to overconfidence during training or fine-tuning. In this work, we observe a consistent phenomenon across different models that the early model is better calibrated, while later training or fine-tuning yields marginal accuracy gains but substantially increases calibration errors. Our analysis suggests that the early model retains uncertainty awareness in both its predictions and features, which is gradually lost with continued training. To avoid training away this uncertainty awareness, we propose \textbf{EUA-Cal}, a novel method that exploits the \textbf{E}arly model as an \textbf{U}ncertainty \textbf{A}nchor for \textbf{Cal}ibration. EUA-Cal introduces early prediction regularization to preserve early predictive uncertainty and prototype structure regularization to exploit uncertainty reflected in the early feature space, jointly mitigating overconfidence. Extensive experiments on image classification and multiple-choice question answering across eight diverse models demonstrate that EUA-Cal outperforms state-of-the-art calibration methods.
الكلمات المفتاحية
الموضوع
بيانات النشر
- المجلة
- غير متاح
- وصول مفتوح
- وصول مفتوح أخضر
اقتبس هذه المقالة
APA 7
Xie, Y., Tang, J., Hua, Z., Ma, Y., Qin, S., Hou, Y., Liu, H., Hou, J., & Jia, Y. (2026). Do Not Train Away Uncertainty: Early Uncertainty Anchored Calibration. https://omanscience.com/ar/articles/do-not-train-away-uncertainty-early-uncertainty-anchored-calibration
MLA 9
Xie, Yutong, et al. "Do Not Train Away Uncertainty: Early Uncertainty Anchored Calibration." https://omanscience.com/ar/articles/do-not-train-away-uncertainty-early-uncertainty-anchored-calibration.
شيكاغو (المؤلف–التاريخ)
Xie, Yutong, Jiawei Tang, Zhenglin Hua, Yuxiang Ma, Si Qin, Yaxin Hou, Hui Liu, Junhui Hou, and Yuheng Jia. 2026. "Do Not Train Away Uncertainty: Early Uncertainty Anchored Calibration." https://omanscience.com/ar/articles/do-not-train-away-uncertainty-early-uncertainty-anchored-calibration.
هارفارد
Xie, Y., Tang, J., Hua, Z., Ma, Y., Qin, S., Hou, Y., Liu, H., Hou, J. and Jia, Y. (2026) 'Do Not Train Away Uncertainty: Early Uncertainty Anchored Calibration', Available at: https://omanscience.com/ar/articles/do-not-train-away-uncertainty-early-uncertainty-anchored-calibration.
فانكوفر
Xie Y, Tang J, Hua Z, Ma Y, Qin S, Hou Y, et al. Do Not Train Away Uncertainty: Early Uncertainty Anchored Calibration. https://omanscience.com/ar/articles/do-not-train-away-uncertainty-early-uncertainty-anchored-calibration
IEEE
Y. Xie, J. Tang, Z. Hua, Y. Ma, S. Qin, Y. Hou, H. Liu, J. Hou, and Y. Jia, "Do Not Train Away Uncertainty: Early Uncertainty Anchored Calibration," https://omanscience.com/ar/articles/do-not-train-away-uncertainty-early-uncertainty-anchored-calibration.