wolfiesch / omp-best-of

Best-of-N coding agents with LLM-as-a-Verifier selection for Oh My Pi: isolated git worktrees, verifier-ranked trajectories, winner-only patch application

marmotlab / RL2-VLA

RL²-VLA: Adaptive RL Latent Compositional Steering with Test-Time Scaling for Vision-Language-Action Models

changdaeoh / sharpening-tax

Official codebase for the paper "Sharpening Tax in Post-Training"