LeRobot ACT

Community implementation

LeRobot implementation of the Action Chunking Transformer policy.

View full model description

Research, evaluation and robot-specific training of Action Chunking Transformer policies.

Papers & reports (1)

Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardware

Paper available

arXiv: 2304.13705
Authors & affiliations (4)

Author order: Original order

  1. Tony Z. Zhao

    Not provided

  2. Vikash Kumar

    Not provided

  3. Sergey Levine

    Not provided

  4. Chelsea Finn

    Not provided

First published
2023-04-23
Latest paper revision
v1 · 2023-04-23
Venue
RSS 2023

Covered model versions

  • act · LeRobot ACT

    Foundational ACT method implemented by LeRobot; not an exact LeRobot implementation release report or a benchmark for an unspecified trained checkpoint.

Paper ↗PDF ↗

How to access and use

Public code resources
Public access

Training / fine-tuning code

Access
Public access

Verified on · 2026-10-27

Release information

SourceVerified on · 2026-10-23

Checks before deployment

Requires task- and embodiment-specific data, configuration and validation. Published examples and simulated results do not establish safe operation on a physical robot. Available weights or code do not establish plug-and-play robot compatibility. Check the required configuration, data and interfaces in the source documentation.

Deployment and usage resources

Runtime and frameworks
LeRobot / PyTorch
Fine-tuning requirements
LeRobot integration: use the linked policy-specific setup and training instructions; compatibility with every upstream checkpoint is not implied.

Architecture and inputs / outputs

Model type
Action policy
Backbone
Transformer encoder-decoder with action chunking
Input
RGB images / Robot state
Output
Continuous joint-action chunks

Scope and training

Continuous joint-action chunks

Functional purpose

Action & control

Record identity

Community implementation

Pipeline task

Robotics

Derived models

Kite Trossen Cup Real ACT ↗Derivative

Kite Trossen Sim ACT Cup Baseline · 30k ↗Derivative

Kite Trossen Cup Sim Baseline ACT ↗Derivative

kiteml/trossen-bin-picking-sim-act-cup-all-90k ↗Derivative

Kite Trossen Sim Cup All ACT 30k ↗Derivative

ME297SJSU2026 SO-101 v3 DAgger3 ↗Derivative

LeRobot ACT · Tridex 3Cam ResNet 100K 18_11_2 ↗Derivative

  • Functional purpose · Record identity · Pipeline task · Continuous joint-action chunks · continuous: LeRobot act documentation

Related robots

Full parameters and verification notes (model card)

LeRobot implements a task-trained visuomotor policy: multi-view RGB observations, current joint positions and a latent style variable feed a transformer encoder/decoder that predicts action chunks. The latent is set to zero at inference. The rollout task-text argument is optional for ACT; English documentation does not establish language-conditioned control. The April 23, 2023 paper introduces the original ACT method, not the release date of this LeRobot implementation. Original ACT software is MIT; the LeRobot implementation is Apache-2.0. Neither code license is used to infer permissions for an unspecified trained checkpoint. No universal pretrained checkpoint, exact inference API, or current-artifact benchmark is asserted here.

Submit a source or correction