Robbyant/lingbot-vla

A Pragmatic VLA Foundation Model

What it solves

LingBot-VLA is a Vision-Language-Action (VLA) foundation model designed to provide a pragmatic, high-performance approach to robotic control. It addresses the challenge of creating models that can generalize across different robot configurations and perform tasks with high success rates in both simulation and real-world environments.

How it works

The model is pre-trained on a massive dataset of 20,000 hours of real-world data from nine different dual-arm robot configurations. It is available in two versions: a standard version and a depth-distilled version that incorporates depth information to improve performance. The codebase is optimized for training efficiency, offering a 1.5x to 2.8x speedup over existing VLA codebases and reduced GPU memory usage.

Who it’s for

This project is intended for robotics researchers and developers who want to deploy VLA models on physical robots or in simulations (such as RoboTwin 2.0), and those looking for efficient post-training pipelines for custom robot data.

Highlights

  • Large-scale Pre-training: Trained on 20,000 hours of data across 9 robot configurations.
  • High Efficiency: Significant training speedups and reduced GPU memory consumption.
  • Depth-Aware Options: Provides both depth-free and depth-distilled model checkpoints.
  • Strong Benchmarks: Achieves state-of-the-art results on GM-100 and RoboTwin 2.0 benchmarks.
  • LeRobot Integration: Fully compatible with LeRobot v3.0 for data handling and deployment.

Related

  • Project
    Robbyant/lingbot-vla-v2

    LingBot‑VLA 2.0 is an open‑source Vision‑Language‑Action foundation model for robots. It unifies many robot embodiments into a 55‑dim action vector, uses MoE layers and dual‑query distillation (depth + future video) to learn predictive dynamics, and is trained on ~60 k h of robot and egocentric video data. The repo provides conda‑based installation, scripts to download 6 B‑parameter checkpoints, tools for data preparation, post‑training on custom datasets, and evaluation on the RoboTwin benchmark and real robots. Reported results show noticeable gains over the 1.0 version on bimanual and mobile manipulation tasks.

  • Project
    XiaomiRobotics/Xiaomi-Robotics-1

    Xiaomi-Robotics-1 is a Vision-Language-Action (VLA) foundation model trained on 100K+ hours of trajectories to enable general-purpose mobile manipulation and efficient task adaptation.

  • Project
    open-gigaai/giga-brain-0

    GigaBrain‑0.7 is an open‑source, large‑scale vision‑language‑action model for robots. It is pretrained on 37 k h of heterogeneous embodied data, supports eight robot embodiments, and ships with training scripts, benchmark suites (VLM‑Benchmark, RoboColiseum, RoboTwin 2.0, EBench), and ready‑to‑run inference servers that integrate with ROS.

  • Project
    sii-research/tau-0-vla

    A hierarchical robot foundation model for long-horizon manipulation that combines a high-level planner with a world-model-guided search and a generalist low-level execution policy.

  • Project
    dexmal/dexbotic

    A PyTorch-based development toolbox for Vision-Language-Action (VLA) models that streamlines pretraining, fine-tuning, and evaluation for embodied intelligence.