Hairong Qi

Measuring Structured Predictability in Neural Training Dynamics: A Cross-Regime Study

This work combines three complementary probe families, displacement-direction, subspace-residual, and predictor-based probes, with convention-aware, null-calibrated group-level readouts, and applies them to multi-pass vision training on CIFAR and public Pythia pretraining checkpoints, indicating that these measurements capture intrinsic trajectory structure, while probe differences distinguish complementary forms of temporal organization.