robocurve
Popular repositories Loading
-
inspect-robots
inspect-robots PublicOpen source evals for physical AI. Run any LLM/VLA on any arm/humanoid against any real/sim benchmark.
-
worldevals
worldevals PublicA curated catalog of VLA / physical-AI benchmarks, each runnable on real robots or sims via Inspect Robots. (The Inspect Evals for robotics.)
-
inspect-robots-yam
inspect-robots-yam PublicInspect Robots adapters for I2RT YAM bimanual arms driven by MolmoAct2
-
stationerybench
stationerybench PublicFive bimanual desk-stationery manipulation tasks for VLA models and LLM agents, built on Inspect Robots
Python 4
-
worldpolicies
worldpolicies PublicRobocurve's registry of published robot-policy (VLA) checkpoints, indexed with eval numbers and validated by CI
-
clapboardbench
clapboardbench PublicReal-robot eval traces: bimanual YAM arms operating a film clapperboard, driven turn by turn by frontier VLMs (Claude Opus, Gemini Robotics-ER). Browsable run pages with the frames each model saw.
HTML 3
Repositories
- inspect-robots Public
Open source evals for physical AI. Run any LLM/VLA on any arm/humanoid against any real/sim benchmark.
-
- stationerybench Public
Five bimanual desk-stationery manipulation tasks for VLA models and LLM agents, built on Inspect Robots
- worldevals Public
A curated catalog of VLA / physical-AI benchmarks, each runnable on real robots or sims via Inspect Robots. (The Inspect Evals for robotics.)
- inspect-robots-so101 Public
Inspect Robots adapters for LeRobot SO-ARM followers (SO-100/SO-101) driven by LeRobot policies
- llm-token-speed Public
- clapboardbench Public
Real-robot eval traces: bimanual YAM arms operating a film clapperboard, driven turn by turn by frontier VLMs (Claude Opus, Gemini Robotics-ER). Browsable run pages with the frames each model saw.
- dreamzero-yam Public
LoRA post-train of NVIDIA DreamZero (world-action model) on the MolmoAct2 BimanualYAM split, trained on Modal; three-model-family comparison vs pi0.5 and GR00T N1.7
- worldpolicies Public
Robocurve's registry of published robot-policy (VLA) checkpoints, indexed with eval numbers and validated by CI