Model family

GAE (General Action Expert)

Participating organizations

Version scope and roles for these associations await verification.

View evidence

GAE (General Action Expert) is a Westlake Robotics and Westlake University framework for real-time humanoid whole-body teleoperation. A generator–executor training design and latency-conditioned anticipation let an executor track human motion goals with robot-state feedback. Separate commercial GAE client integration is documented through a local WebSocket API. The pinned publisher README requires purchased software and a paired USB dongle/key. Current regional download metadata lists softwareV1.2.0; this is not a policy-weight revision or a public hosted model service. Equality with the research executor’s weights is unverified.

Overview

Model type
Robot foundation model
Listed versions
GAE research 2026

Model history

September 2026

GAE paper details generator–executor learning and anticipation trade-offs

New resourcesOfficial announcement

ArXiv v1 describes each approximately852M policy, PPO generator–executor training, simulation tracking results and separately scoped real demonstrations. Anticipation0–5steps at50Hz trades tracking precision against0–100ms configured lead time; it is not measured end-to-end latency.

Previously reported

The author project already documented the2025 public teleoperation demonstration, without the current paper’s exact training and quantitative evaluation record.

This announcement

ArXiv v1 describes each approximately852M policy, PPO generator–executor training, simulation tracking results and separately scoped real demonstrations. Anticipation0–5steps at50Hz trades tracking precision against0–100ms configured lead time; it is not measured end-to-end latency.

Ordinary: technical documentation of an existing model, without verified new public weights, training-code access or a major independently established capability release. Some errors worsen against SONIC, and model/runtime identity with licensed commercial software is unconfirmed. Admission score95 does not determine this editorial decision.

September 2025

GAE publicly demonstrates general humanoid whole-body teleoperation

New modelOfficial announcement

The author project dates the initial GAE demonstration to September29,2025 and shows a humanoid following diverse operator motions as a physical avatar. This establishes public model/demonstration evidence, not open weights or independent performance certification.

Previously reported

The reviewed author record does not establish an earlier public GAE model demonstration or a prior downloadable GAE checkpoint.

This announcement

The author project dates the initial GAE demonstration to September29,2025 and shows a humanoid following diverse operator motions as a physical avatar. This establishes public model/demonstration evidence, not open weights or independent performance certification.

Featured: a distinct general whole-body teleoperation model changes the motion-control capability options users can evaluate. Do not transfer the later paper’s852M parameters or58 results to an unverified identical2025 checkpoint. Human-guided control is not autonomous household planning; no world-first claim is independently confirmed.

View all important updates →

Versions

1

Papers & reports (1)

GAE: General Action Expert for Real-Time Humanoid Teleoperation

Paper available

arXiv: 2609.34233DOI: 10.48550/arxiv.2609.34233
Authors & affiliations (9)

Author order: Original order

  1. Yuefan Wang

    Westlake Robotics · Westlake University

  2. Huaicheng Zhou

    Westlake Robotics

  3. Xiao He

    Westlake Robotics

  4. Zhijie He

    Westlake Robotics

  5. Mingchuan Yang

    Westlake Robotics

  6. Huayi Zhang

    Westlake Robotics

  7. Li Chai

    Westlake University

  8. Jinxin LiuCorresponding author

    Westlake Robotics

  9. Donglin WangCorresponding author

    Westlake Robotics · Westlake University

First published
2026-09-28
Latest paper revision
v1 · 2026-09-28
Venue
Not provided

Covered model versions

  • gae · GAE research 2026

    GAE 2026 research framework and executor/generator policies; simulation evaluations and separately described human-operated robot demonstrations, not autonomous WR1 task planning or a public checkpoint.

Paper ↗PDF ↗

Official resources

Submit a source or correction