Team Dexterity: training pi0.5 (openpi, JAX) for the 2026 BEHAVIOR Challenge -- code, step-by-step guide, model and results
-
Updated
Nov 15, 2026 - Python
Team Dexterity: training pi0.5 (openpi, JAX) for the 2026 BEHAVIOR Challenge -- code, step-by-step guide, model and results
100-task BEHAVIOR-1K (2026) meta-training pipeline for a pi-0.5 VLA policy on 8x B300. Extends the 50-task BEHAVIOR-Challenge-winning checkpoint to 100 activities.
Development-selected routing of public robot policies for BEHAVIOR-1K (Team KANR, 2026 BEHAVIOR Challenge): selector, routing, scoring, fail-stop serving, data and paper
Evidence-backed, model-agnostic runtime contracts for physical agents
Closed-loop benchmark of 9 VLA policies (GR00T N1.7, pi0/openpi, LoRA vs full fine-tuning) on BEHAVIOR-1K. Imitation learning / behavioral cloning with diffusion action heads in PyTorch, plus a reproducible eval harness, unattended training automation, and raw metrics.
HiPhys-RSI — 1st place, 2026 BEHAVIOR Challenge (BEHAVIOR-1K): Q 0.683, 48.2% success on all 100 tasks. Project page.
To associate your repository with the behavior-1k topic, visit your repo's landing page and select "manage topics."