Causation

Degradation of Model Calibration from Post-Training Alignment

Post-training alignment (such as PPO reinforcement learning) can significantly degrade the calibration of large language models. In pre-trained models such as GPT-4, predicted probabilities (logprobs) across multiple-choice options track actual task accuracy closely, yielding near-perfect calibration with an Expected Calibration Error (ECE) of $0.007on benchmarks like MMLU. However, subsequent post-training alignment impairs this correspondence, substantially inflating calibration error to an ECE of $0.074 and causing the model's confidence to diverge from its true probability of being correct.

0

1

Updated 2026-09-07

Tags

Prep Sessions

Transformer Architecture and Large Language Model Capabilities @ University of Michigan - Ann Arbor

Ch.3 Model Alignment and Safety - Transformer Architecture and Large Language Model Capabilities @ University of Michigan - Ann Arbor

Model Calibration and Confidence Degradation - Transformer Architecture and Large Language Model Capabilities @ University of Michigan - Ann Arbor