Armnet releases ArmnetBench v0.1
Armnet released ArmnetBench v0.1, a parallel real-world benchmark for manipulation policies run on its low-cost SO-101 robot-arm farm.
Evidence notes
- The benchmark evaluates seven policy families across eight single-arm and four bimanual tasks.
- Its core release contains 2,518 policy rollouts and 600 human-reference demonstrations with three-way outcome labels.
- Armnet published the benchmark report, datasets and repository for reproducible inspection and reuse.
Company context
Armnet operates an open, cloud-connected robot-arm farm for physical-AI researchers. Developers package policies and other workloads in Docker, submit them with Python and run them on ready-to-use physical robot arms.