pi05_FAST_aloha_pick_up_ES

ALOHA pick-up, ES (baseline) checkpoint from Fine-Tuning VLAs with Self-Demonstrated Generative Control for Multi-Task Manipulation.

Paper · Project page · Code

• Model: π0.5 with FAST tokens, fine-tuned from pi05_base
• Training data: Expert demos only (aloha_pick_up_red_green_cubes_expert).
• Step: 9000 (params/ and assets/ only, no optimizer state)
• Norm stats: assets/trossen/norm_stats.json

Usage

From the root of the code repo:

hf download prachigarg23/pi05_FAST_aloha_pick_up_ES --local-dir checkpoints/pi05_FAST_aloha_pick_up_ES
python scripts/serve_policy.py policy:checkpoint --policy.config=pi05_FAST_aloha_pick_up_ES \
    --policy.dir=checkpoints/pi05_FAST_aloha_pick_up_ES

See the ALOHA README for running it on the robot.

Citation

@misc{garg2026finetuningvlasselfdemonstratedgenerative,
      title={Fine-Tuning VLAs with Self-Demonstrated Generative Control for Multi-Task Manipulation},
      author={Prachi Garg and Steve Xing and Prahit Yaugand and Saurabh Gupta and Derek Hoiem},
      year={2026},
      eprint={2608.19490},
      archivePrefix={arXiv},
      primaryClass={cs.RO},
      url={https://arxiv.org/abs/2608.19490},
}