You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Post-training a small LLM for table numerical reasoning: execution-feedback GRPO + PRM-lite process reward + best-of-n test-time compute. Runs on a single 24GB GPU.