VLA-REPLICA benchmark accepted at NeurIPS 2026, adds MolmoAct 2 and GR00T N1.7
YuXiang_IRVL · x · 2026-09-27
VLA-REPLICA accepted to NeurIPS 2026 E&D Track
- What it is: a low-cost, reproducible real-world benchmark for evaluating VLA models, addressing the lack of standardized evaluation.
- How: standardized hardware setup and scene initialization make it easily reproducible across labs and institutions.
- Updates: leaderboard now includes MolmoAct 2 and GR00T N1.7; all models evaluated over 3 rounds in both ID and OOD settings with error intervals.
More from Embodied
- Tesla Optimus scales to hundreds per week but hands remain a bottleneck — AIFlow_ML · 2026-09-27
- $800 litter-picking robot MOSS hits V0.3, open-source V0.4 already printing — AIFlow_ML · 2026-09-27
- GPT-6 Astra Robot Demos Look Great on X, But Real-Time Footage Shows Painful Slowness — DJiafei · 2026-09-27
- Humanoid Robotics Debate: Claims US Firms Fake Demos While Chinese Companies Stay Legit — cixliv · 2026-09-27
- Robots crack whips with zero learning: motion capture retargeting enables dynamic manipulation — ATTlKA · 2026-09-27
- Asimov open-sources humanoid locomotion policy; non-technical user fine-tunes fall recovery on a single 4090 — freelerobot · 2026-09-27