Learning Shrinks the Hard Tail: Training-Dependent Inference Scaling in a Solvable Linear Model

Add code
Jan 07, 2026

Share this with someone who'll enjoy it:

View paper onarxiv icon

Share this with someone who'll enjoy it: