YijiaFan/CaPtain-4B
Image-Text-to-Text • 5B • Updated • 113 • 2
CaPtain: A Pipeline for Post-Training Code-as-Policy Robot Agents — models, SFT data and artifacts.
Note Main model (SFT + RL, round 300)
Note SFT-only model before RL
Note SFT trajectories
Note Critic weights, configs, eval start states, example episodes