SO-101 Bimanual Baseline (GR00T-N1.7-3B fine-tune)
Baseline pair for "Act with Intent: Brain-to-Spine Intention Distillation" (INDI). Vanilla GR00T-N1.7-3B fine-tune (no IntD stream), 60k steps (GBS 128) on the same 400 bimanual SO-101 teleop demos (4 tasks × 100) as the IntD model.
- Action space: relative arm / absolute gripper joint actions (official GR00T SO-100 convention), chunk 32, executed with receding horizon 8.
- Serve with the intd-vla fork
(
gr00t/eval/run_gr00t_server.py --embodiment-tag SO101_BIMANUAL).
Real-world average success (n=25/cell): 71 (ID) / 62 (held-out) / 53 (distractors) vs IntD 76/67/60 — see leesangoh/intd-so101-bimanual.
- Downloads last month
- 20
Model tree for leesangoh/intd-so101-bimanual-baseline
Base model
nvidia/GR00T-N1.7-3B