DexRobotics open-sources a 50-episode SO101 robot fine-tuning workflow

AdinaYakup · x · 2026-07-29

DexRobotics open-sourced an SO101 pick-and-place SFT checkpoint and the full LoRA fine-tuning workflow for DM0.5.

• The demo fine-tunes DM0.5 on a real SO101 cube-picking task.
• Training used only 50 episodes, and the dataset focused on an orange cube.
• After fine-tuning, the robot can pick the orange cube and place it in a plate.
• DexRobotics says the model also generalizes to unseen cube colors, shifted plate positions, and disturbances during grasping and placing.
• The release includes the pretrained model, the SO101 SFT model, the dataset, and a guide, with the broader goal of helping low-cost robots learn new skills from small datasets.

Original post →

More from Embodied

• HUG learns robot grasping from 1 million human handshapes and beats baselines by 34% — LerrelPinto · 2026-07-29
• Directory of US Quick-Turn Hardware Suppliers Amid FCC Robot Bans — mattfreed · 2026-07-29
• Khora proposes a multi-agent world model that scales linearly instead of quadratically — siyuanhuang95 · 2026-07-29
• Nature Communications: Spatial Network Principles Underlying Neural Locomotion — plopesresearch · 2026-07-29
• China shows a wall-climbing embodied robot as BYD, Honor and Arm push new hardware — 创业邦 · 2026-07-29
• Seattle demo shows a rideable robot and autonomy for scooters and bikes — Scobleizer · 2026-07-29

Embodied channel →