AI RESEARCH

On the Unreasonable Effectiveness of Last-layer Retraining

arXiv CS.LG

ArXi:2512.01766v2 Announce Type: replace Last-layer re