Does ACT run on SO-101?

Capabilities

Evidence 4 sources; the cell shows the highest tier

1. T1 · upstream documentedofficial-docssource 2026-08-03verified 2026-10-24 (1d ago)lerobot 0.6.1

   LeRobot v0.6.1 ACT guide evaluates a trained ACT policy on the real robot with lerobot-rollout --robot.type=so101_follower.

   “lerobot-rollout \ --strategy.type=base \ --policy.path=${HF_USER}/act_policy \ --robot.type=so101_follower \”

   https://github.com/huggingface/lerobot/blob/v0.6.1/docs/source/act.mdx
2. T1 · upstream documentedofficial-docssource 2026-08-03verified 2026-10-24 (1d ago)lerobot 0.6.1

   The v0.6.1 real-robot imitation-learning tutorial records on so101_follower and trains ACT (--policy.type=act) on the resulting so101_test dataset.

   “lerobot-train \ --dataset.repo_id=${HF_USER}/so101_test \ --policy.type=act \ --output_dir=outputs/train/act_so101_test \”

   https://github.com/huggingface/lerobot/blob/v0.6.1/docs/source/il_robots.mdx
3. T1 · upstream documentedofficial-docssource 2026-08-03verified 2026-10-24 (1d ago)lerobot 0.6.1

   The v0.6.1 imitation-learning tutorial teleoperates and records datasets with an SO101 leader/follower pair.

   “In this example, we’ll demonstrate how to teleoperate the SO101 robot.”

   https://github.com/huggingface/lerobot/blob/v0.6.1/docs/source/il_robots.mdx
4. T1 · upstream documentedpapersource 2025-06-02verified 2026-10-24 (1d ago)

   SmolVLA paper, Table 4: real-world SO101 Pick-Place-Lego benchmark, ACT vs SmolVLA, in- and out-of-distribution.

   “Similarly, on SO101 (see Table 4), SmolVLA surpasses ACT in both in-distribution and out-of-distribution (OOD) settings.”

   https://arxiv.org/html/2506.01844v1