🏆 Leaderboard   ·   🤖 Models   ·   📖 Protocol   ·   📤 Submit

Install and run

git clone https://github.com/MotionChallenge-Project/MotionChallenge-Core.git
cd MotionChallenge-Core
uv sync

Run a checkpoint from a demo adapter with uv run evaluate.py. Each adapter pins MotionChallenge-Core and produces verified results using the shared protocol.

Demo Model adapters

Model	Repository
GRIT	Evaluation repository
SONIC v1.1	Evaluation repository
HoloMotion v1.4.1	Evaluation repository
PMT RGMT	Evaluation repository

Clone an adapter with --recurse-submodules, then run uv sync and uv run evaluate.py.

Evaluation

Full-motion success, completion, joint error, and body-position error with XY / XY+yaw / no alignment. Frozen physics and protocol; hashes and pose traces support independent recomputation. Results, data, and weights stay outside Git. Each result identifies its dataset, robot, checkpoint and protocol.

Protocol · Add a model · Add a benchmark · Batch evaluation · Code layout · Website · Project

Apache-2.0. Third-party notices.

All controllers use ideal proprioception without global position feedback.

Citation

If you use MotionChallenge in your research, please cite:

@misc{motionchallenge2026,
  author       = {{MotionChallenge Team}},
  title        = {{MotionChallenge}: An Open, Reproducible Benchmark for Humanoid Motion Tracking},
  year         = {2026},
  url          = {https://motionchallenge-project.github.io/motionchallenge/}
}

Download BibTeX.