Meta researcher: do as much SFT as possible before touching RL, or you're wasting money
DimitrisPapail · x · 2026-11-06
Meta researcher Dimitris Papail offers a training take: "The bitter lesson is you have to do as much SFT as possible before you touch any RL. Unless you like spending money for no reason."
The claim frames SFT quality as the dominant cost-efficiency lever, arguing RL investment pays off poorly if the supervised base isn't maximized first.
More from Research
- Kording Lab amplifies thread: objective functions as concise descriptions of what neural nets learn — KordingLab · 2026-11-06
- LLM peer review is here to stay, researchers say — but 'excuses to reject' era must end — examachine · 2026-11-06
- UChicago names 31 researchers to 2026 Rising Stars in Data Science, with focus on world models — cindy_x_wu · 2026-11-06
- Surge researcher praises Tinker: parallel RL experiments without GPU management — echen · 2026-11-06
- Nemotron 3.5 Super + Surge RL matches GPT-5.6 Terra across 12 benchmarks — echen · 2026-11-06
- BMVA symposium on computer vision interpretability opens CFP, London, Jan 2027 — HildeKuehne · 2026-11-06