TY  - RPRT
TI  - When does a network's training history predict its future learning better than its current state? Evidence from a response probe and a forecasting screen
AU  - Martin Hofmann
AU  - Patrick Mäder
PY  - 2026
UR  - https://arxiv.org/abs/2610.09621
ID  - 2610.09621
ER  -