Do Not Train Away Uncertainty: Early Uncertainty Anchored Calibration
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.
Key points
- Deep neural networks, including large language models, have achieved remarkable performance across various tasks.
- 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.
- To avoid training away this uncertainty awareness, we propose EUA-Cal, a novel method that exploits the Early model as an Uncertainty Anchor for Calibration.
- 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.
Sources (1)
- [1]Do Not Train Away Uncertainty: Early Uncertainty Anchored CalibrationarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 8, 02:34 PM
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.
Deep neural networks, including large language models, have achieved remarkable performance across various tasks.
Extractive summary: sentences quoted from the sources.
Before this
- Oct 8, 2026Beyond Imitation: A Framework and Benchmark for LLM-Assisted Peer Review
- Oct 8, 2026SuperNav: An Agentic Navigation System for Any Task in Any Scene
- Oct 8, 2026VibeEdit: Image Editing with Canvas Instructions
- Oct 8, 2026SpaceCast-Bench: Evaluating Predictive Spatial Reasoning in Vision-Language Models
- Oct 7, 2026Iris-3B: Going Beyond the Latent with Pixel-Space Diffusion Training, Conversion and Fine-Tuning
- Oct 7, 2026Q-Learning with Scalar Adjoint Matching