SFT+RL pipeline choice makes the lazy-distillation theory unlikely, argues dev
JoshPurtell · x · 2026-09-03
Follow-up in JoshPurtell's distillation debate thread:
- If they lazily distilled from sol, the only rationale would be not wanting to spin up Kimi on openrouter — but that doesn't hold either
- Distilling would mean not adding an SFT stage to the RL pipeline (especially with good starting performance), and SFT+RL is harder than RL-only
- Conclusion: possible but not 80%
Complements the earlier post: the model's gains likely come from RL, not distillation.
More from Models
- Models can detect read-only probes with just last-two-layer SFT — Sauers_ · 2026-10-18
- Paying Google AI subscriber blocked: custom-domain Workspace accounts can't access AI Studio — ChrisRackauckas · 2026-10-18
- Running Ling 3.1 Flash (560B MoE, 25B Active) Across 4 DGX Sparks via llama.cpp RPC — BerryBlush_Rogers691 · 2026-10-18
- Users press OpenAI on long-delayed Adult Mode a year after it was promised — ___Patrice___ · 2026-10-18
- Polymarket opens betting market on who releases the next Gemini Argon model — Polymarket · 2026-10-18
- Grok matter-of-factly lists which groups user was racist against, exempting Africans and Latin Americans — teortaxesTex · 2026-10-18