New Study Reexamines Whether RL Generalizes While SFT Memorizes
An arXiv paper revisits the claim that reinforcement learning with verifiable rewards generalizes while supervised fine-tuning memorizes, focusing on the role of SFT data in that contrast.
Read the full story at arXiv cs.LG ↗
Timeline · 1 report
- 2026-11-13 05:00 · arXiv cs.LG
Rethinking "RL Generalizes, SFT Memorizes": The Role of SFT Data