AI RESEARCH
On the Unreasonable Effectiveness of Last-layer Retraining
arXiv CS.LG
•
ArXi:2512.01766v2 Announce Type: replace Last-layer re