ScaleBFM M

Author project · Checkpoint · Provenance

Install and run

git clone --recurse-submodules https://github.com/MotionChallenge-Project/MotionChallenge-scalebfm-m.git
cd MotionChallenge-scalebfm-m
uv sync
uv run evaluate.py

Produces success, completion and six MPBPE values in a verified result.json. MuJoCo 3.17 CPU batch; --jobs sets parallelism. Missing data, weights and pinned author inference files are downloaded and hash-checked; results and traces stay outside Git.

Inference validation · Native parity report

Observations

Ideal proprioception, without robot world-position or contact feedback. Checkpoint requirements · Rules.

Initialize root history from the first reference root position instead of measured robot position, which is unavailable under the benchmark observation rules. Then use the author mode 7 reference-forcing inference with the original checkpoint; no robot translation is measured or integrated.