Records · kv:2610.00019v1

Reproduction of huggingface/lerobot: Diffusion Policy on PushT

K-Veritas Team

Submitted by K-Veritas Team

Published 2026-10-08

Independent reproduction · Artifact evaluation · Robotics

PaperReproduction of: original code

Abstract

Trained LeRobot's Diffusion Policy on the lerobot/pusht dataset at the pinned commit for 20,000 steps (batch 64, seed 1000, default diffusion settings), a tenth of the 200,000-step budget used for the published checkpoint, and evaluated it in the gym-pusht simulator on 50 episodes at 10,000 and 20,000 steps (success rate and maximum coverage reward). Evaluation ran in-process (use_async_envs=false) because the asynchronous evaluation workers could not load the simulator. One NVIDIA RTX 5060 Ti.

Sealed report

Cite this record

@misc{kveritas261000019v1,
  title={Reproduction of huggingface/lerobot: Diffusion Policy on PushT},
  author={K-Veritas Team},
  year={2026},
  howpublished={K-Veritas Records},
  note={kv:2610.00019v1},
  url={https://kveritas.org/records/2610.00019v1},
}

References

1. [1] Mamadou K. Keita and Christopher Homan. Computer Science Conferences Should Require Nonrepudiable Experimental Results. NeurIPS 2026 Position Paper Track. arXiv:2605.08586.

Sealed with K-Veritas [1]. Listed, not endorsed. A record shows these results came from the sealed code, unchanged. Whether the work is sound is for the reader to judge.