τ₀-VLA is a hierarchical robot foundation model for long-horizon manipulation that uses world-model-guided test-time computation to search over subtask alternatives when additional reasoning is needed.

Stars

155

7-day growth

No data

Forks

8

Open issues

0

License

Apache-2.0

Last updated

2026-07-27

AI repository intelligence
FR-AI / ANALYSIS

Why it is worth attention

It introduces a memory-augmented high-level policy that leverages a world model for test-time reasoning, combined with a large-scale low-level policy trained on 40,115 hours of heterogeneous real-world robot data, and is fully open-sourced.

Who it is for

  • Robotics researchers studying long-horizon manipulation
  • AI/ML engineers developing robot foundation models
  • Practitioners deploying post-training on specific robot embodiments
  • Researchers interested in test-time computation for robot reasoning

Use cases

  • Long-horizon manipulation tasks requiring hierarchical planning
  • Post-training a generalist robot policy on custom datasets or robots
  • Benchmarking hierarchical reasoning and world-model-guided policies
  • Open-loop evaluation of joint-control robot policies

Strengths

  • Hierarchical architecture with world-model-guided test-time computation enables adaptive reasoning
  • Low-level policy trained on 40,115 hours of diverse real-world robot data with multimodal co-training
  • Open-source weights, code, and example data under Apache 2.0 license
  • Flexible post-training support for custom datasets and robot embodiments

Considerations

  • Serving only supports joint-control checkpoints; end-effector serving is not included in this release
  • Evaluation is limited to open-loop only; closed-loop or real-time deployment not yet documented
  • Test-time computation may increase inference latency for complex reasoning steps

README quick start

Installation

Description

This repo is the official implementation of "τ0-VLA: a Hierarchical Robot Foundation Model with World-Model-Guided Test-Time Computation".

Related repositories

Similar projects matched by category, topics, and programming language.

lopopolo
Featured
lopopolo GitHub avatar

harness-engineering

Harness Engineering is a methodology for improving coding agent outputs by carefully crafting the environment around them—providing curated context, tools, and executable constraints that encode an organization’s nonfunctional requirements and cumulative lessons.

AI & Machine LearningAI Agents
2,390
slvDev
Featured
slvDev GitHub avatar

esp32-ai

A 28.9 million parameter language model runs on an $8 ESP32-S3 microcontroller entirely on-device, generating simple stories at about 9.5 tokens per second.

AI & Machine LearningLarge Language Models
1,960
littledivy
Featured
littledivy GitHub avatar

mimic

mimic captures traffic from any iOS or web app and automatically generates a Python client library that lets you call the app's API like a regular library.

AI & Machine Learning
1,482