[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-72371":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":10,"language":11,"languages":10,"totalLinesOfCode":10,"stars":12,"forks":13,"watchers":14,"openIssues":14,"contributorsCount":15,"subscribersCount":15,"size":15,"stars1d":16,"stars7d":17,"stars30d":18,"stars90d":15,"forks30d":15,"starsTrendScore":19,"compositeScore":20,"rankGlobal":10,"rankLanguage":10,"license":10,"archived":21,"fork":21,"defaultBranch":22,"hasWiki":21,"hasPages":21,"topics":23,"createdAt":10,"pushedAt":10,"updatedAt":28,"readmeContent":29,"aiSummary":30,"trendingCount":15,"starSnapshotCount":15,"syncStatus":31,"lastSyncTime":32,"discoverSource":33},72371,"AgiBot-World","OpenDriveLab\u002FAgiBot-World","OpenDriveLab","[IROS 2025 Best Paper Award Finalist & IEEE TRO 2026] The Large-scale Manipulation Platform for Scalable and Intelligent Embodied Systems","https:\u002F\u002Fopendrivelab.com\u002FAgiBot-World\u002F",null,"Python",3008,206,36,0,6,9,58,18,28.95,false,"main",[24,25,26,27],"pretraining-for-robotics","robotic-foundation-model","robotic-manipulation","vision-language-action-model","2026-06-12 02:03:02","\u003Cdiv id=\"top\" align=\"center\">\n\n![agibot_world](https:\u002F\u002Fgithub.com\u002Fuser-attachments\u002Fassets\u002Fdf64b543-db82-41ee-adda-799970e8a198)\n\n\u003Ca href=\"https:\u002F\u002Fopendrivelab.com\u002FOpenGO1\u002F\" target=\"_blank\">Research Blog: GO-1 Open-sourcing\u003C\u002Fa> | \u003Ca href=\"https:\u002F\u002Fopendrivelab.com\u002Fblog\u002Fagibot-world\u002F\" target=\"_blank\">Research Blog: AgiBot World Colosseo\u003C\u002Fa> | \u003Ca href=\"https:\u002F\u002Farxiv.org\u002Fabs\u002F2503.06669\" target=\"_blank\">Technical Report\u003C\u002Fa>\n\n\u003Ca href=\"https:\u002F\u002Farxiv.org\u002Fabs\u002F2503.06669\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FarXiv-Paper-\u003Ccolor>\">\u003C\u002Fa> [![Static Badge](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FDataset-grey?style=plastic&logo=huggingface&logoColor=yellow)](https:\u002F\u002Fhuggingface.co\u002Fagibot-world) [![Static Badge](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FProject%20Page-blue?style=plastic)](https:\u002F\u002Fagibot-world.com) [![License](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FLicense-CC_%20_BY--NC--SA_4.0-blue.svg)](https:\u002F\u002Fcreativecommons.org\u002Flicenses\u002Fby-nc-sa\u002F4.0\u002F)\n\u003Ca href=\"https:\u002F\u002Fdocs.google.com\u002Fspreadsheets\u002Fd\u002F1GWMFHYo3UJADS7kkScoJ5ObbQfAFasPuaeC7TJUr1Cc\u002Fedit?usp=sharing\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FDataset-Overview-brightgreen?logo=googleforms\" alt=\"Document Badge\">\u003C\u002Fa> [![Static Badge](https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FModel-grey?style=plastic&logo=huggingface&logoColor=yellow)](https:\u002F\u002Fhuggingface.co\u002Fagibot-world\u002FGO-1)\n\n\u003C\u002Fdiv>\n\nAgiBot World Colosseo is a full-stack large-scale robot learning platform curated for advancing bimanual manipulation in scalable and intelligent embodied systems. It is accompanied by foundation models, benchmarks, and an ecosystem to democratize access to high-quality robot data for the academic community and the industry, paving the path towards the \"ImageNet Moment\" for Embodied AI.\n\nWe have released:\n- **\u003Ca href=\"https:\u002F\u002Fhuggingface.co\u002Fagibot-world\u002FGO-1\" target=\"_blank\">GO-1\u003C\u002Fa>:** Our robotic foundation model pretrained on AgiBot World Dataset\n- **\u003Ca href=\"https:\u002F\u002Fhuggingface.co\u002Fagibot-world\u002FGO-1-Air\" target=\"_blank\">GO-1 Air\u003C\u002Fa>:** GO-1 model without Latent Planner, high-performanced and lightweighted\n- **\u003Ca href=\"https:\u002F\u002Fdocs.google.com\u002Fspreadsheets\u002Fd\u002F1GWMFHYo3UJADS7kkScoJ5ObbQfAFasPuaeC7TJUr1Cc\u002Fedit?usp=sharing\" target=\"_blank\">Task Catalog\u003C\u002Fa>:** Reference sheet outlining the tasks in our dataset, including robot end-effector types, sample action-text descriptions and more\n- **\u003Ca href=\"https:\u002F\u002Fhuggingface.co\u002Fdatasets\u002Fagibot-world\u002FAgiBotWorld-Beta\" target=\"_blank\">AgiBot World Beta\u003C\u002Fa>:** Our complete dataset featuring 1,003,672 trajectories (~43.8T)\n- **\u003Ca href=\"https:\u002F\u002Fhuggingface.co\u002Fdatasets\u002Fagibot-world\u002FAgiBotWorld-Alpha\" target=\"_blank\">AgiBot World Alpha\u003C\u002Fa>:** A curated subset of AgiBot World Beta, containing 92,214 trajectories (~8.5T)\n\n## News📰 \u003Ca name=\"news\">\u003C\u002Fa>\n\n> [!IMPORTANT]\n> 🌟 Stay up to date at [opendrivelab.com](https:\u002F\u002Fopendrivelab.com\u002F#news)!\n\n- **`[2025\u002F09\u002F19]`** 🚀 **Our robotic foundation model GO-1 open-sourced.**\n- **`[2025\u002F03\u002F10]`** 📄 \u003Ca href=\"https:\u002F\u002Fopendrivelab.com\u002Fblog\u002Fagibot-world\u002F\" target=\"_blank\">Research Blog\u003C\u002Fa> and \u003Ca href=\"https:\u002F\u002Farxiv.org\u002Fabs\u002F2503.06669\" target=\"_blank\">Technical Report\u003C\u002Fa> released.\n- **`[2025\u002F03\u002F01]`** Agibot World Beta released.\n- **`[2025\u002F01\u002F03]`** \u003Cspan style=\"color: #B91C1C; font-weight: bold;\">Agibot World Alpha Sample Dataset released.\u003C\u002Fspan>\n- **`[2024\u002F12\u002F30]`** 🤖 Agibot World Alpha released.\n\n## TODO List 📅 \u003Ca name=\"todolist\">\u003C\u002Fa>\n\n- [x] **AgiBot World Alpha**\n- [x] **AgiBot World Beta**\n  - [x] ~1,000,000 trajectories of high-quality robot data \n- [x] **AgiBot World Foundation Model: GO-1**\n  - [x] GO-1 fine-tuning script\n  - [x] GO-1 Air pre-trained checkpoint\n  - [x] GO-1 pre-trained checkpoint\n  - [x] Examples of using GO-1 model\n- [x] **2025 AgiBot World Challenge**\n\n## Key Features 🔑 \u003Ca name=\"keyfeatures\">\u003C\u002Fa>\n\n- **1 million+** trajectories from 100 robots.\n- **100+ 1:1 replicated real-life scenarios** across 5 target domains.\n- **Cutting-edge hardware:** visual tactile sensors \u002F 6-DoF Dexterous hand \u002F mobile dual-arm robots\n- **Wide-spectrum versatile challenging tasks**\n- **General robotic policy pretrained on AgiBot World**\n\n\u003Cdiv style=\"max-width: 100%; overflow-x: auto; margin: 0 auto; !important;\">\n  \u003Ctable style=\"border-collapse: collapse; border-spacing: 0; width: 100%; table-layout: fixed;\">\n    \u003Ctr style=\"border: none;\">\n      \u003Ctd align=\"center\" style=\"border: none; padding: 10px;\">\n        \u003Cimg src=\"assets\u002FContact-rich_manipulation.gif\" alt=\"Contact-rich Manipulation\" width=\"230\" style=\"border-radius: 10px; box-shadow: 0 4px 8px rgba(0, 0, 0, 0.1);\">\n        \u003Cp>\u003Cb>Contact-rich Manipulation\u003C\u002Fb>\u003C\u002Fp>\n      \u003C\u002Ftd>\n      \u003Ctd align=\"center\" style=\"border: none; padding: 10px;\">\n        \u003Cimg src=\"assets\u002FLong-horizon_planning.gif\" alt=\"Long-horizon Planning\" width=\"230\" style=\"border-radius: 10px; box-shadow: 0 4px 8px rgba(0, 0, 0, 0.1);\">\n        \u003Cp>\u003Cb>Long-horizon Planning\u003C\u002Fb>\u003C\u002Fp>\n      \u003C\u002Ftd>\n      \u003Ctd align=\"center\" style=\"border: none; padding: 10px;\">\n        \u003Cimg src=\"assets\u002FMulti-robot_collaboration.gif\" alt=\"Multi-robot Collaboration\" width=\"230\" style=\"border-radius: 10px; box-shadow: 0 4px 8px rgba(0, 0, 0, 0.1);\">\n        \u003Cp>\u003Cb>Multi-robot Collaboration\u003C\u002Fb>\u003C\u002Fp>\n      \u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr style=\"border: none;\">\n      \u003Ctd align=\"center\" style=\"border: none; padding: 10px;\">\n        \u003Cimg src=\"assets\u002Fagilex_fold_shirts.gif\" alt=\"Fold Shirt (AgileX)\" width=\"230\" style=\"border-radius: 10px; box-shadow: 0 4px 8px rgba(0, 0, 0, 0.1);\">\n        \u003Cp>\u003Cb>Fold Shirt (AgileX)\u003C\u002Fb>\u003C\u002Fp>\n      \u003C\u002Ftd>\n      \u003Ctd align=\"center\" style=\"border: none; padding: 10px;\">\n        \u003Cimg src=\"assets\u002Fg1_fold_shirts.gif\" alt=\"Fold Shirt (AgiBot G1)\" width=\"230\" style=\"border-radius: 10px; box-shadow: 0 4px 8px rgba(0, 0, 0, 0.1);\">\n        \u003Cp>\u003Cb>Fold Shirt (AgiBot G1)\u003C\u002Fb>\u003C\u002Fp>\n      \u003C\u002Ftd>\n      \u003Ctd align=\"center\" style=\"border: none; padding: 10px;\">\n        \u003Cimg src=\"assets\u002Ffranka_fold_shirts.gif\" alt=\"Fold Shirt (Dual Franka)\" width=\"230\" style=\"border-radius: 10px; box-shadow: 0 4px 8px rgba(0, 0, 0, 0.1);\">\n        \u003Cp>\u003Cb>Fold Shirt (Dual Franka)\u003C\u002Fb>\u003C\u002Fp>\n      \u003C\u002Ftd>\n    \u003C\u002Ftr>\n  \u003C\u002Ftable>\n\u003C\u002Fdiv>\n\n## Table of Contents\n\n- [News📰 ](#news-)\n- [TODO List 📅 ](#todo-list--)\n- [Key Features 🔑 ](#key-features--)\n- [Table of Contents](#table-of-contents)\n- [Getting started 🔥 ](#getting-started--)\n  - [Installation ](#installation-)\n  - [How to Get Started with Our AgiBot World Data ](#how-to-get-started-with-our-agibot-world-data-)\n    - [Download Datasets ](#download-datasets-)\n    - [Visualize Datasets ](#visualize-datasets-)\n  - [How to Get Started with Our GO-1 Model ](#how-to-get-started-with-our-go-1-model-)\n    - [Requirements ](#requirements-)\n    - [Model Zoo ](#model-zoo-)\n    - [Fine-tuning on Your Own Dataset ](#fine-tuning-on-your-own-dataset-)\n    - [Testing Your Model ](#testing-your-model-)\n    - [More Examples ](#more-examples-)\n- [License and Citation📄   ](#license-and-citation---)\n\n## Getting started 🔥 \u003Ca name=\"gettingstarted\">\u003C\u002Fa>\n\n### Installation \u003Ca name=\"installation\">\u003C\u002Fa>\n\n1. Download our source code:\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002FOpenDriveLab\u002FAgiBot-World.git\ncd AgiBot-World\n```\n\n2. Create a new conda environment:\n```bash\nconda create -n go1 python=3.10 -y\nconda activate go1\n```\n\n3. Install dependencies:\n> This project is built on [LeRobot](https:\u002F\u002Fgithub.com\u002Fhuggingface\u002Flerobot) (**dataset `v2.1`, commit `2b71789`**)  \n> ⚡️ Our environment has been tested with **CUDA 12.4**.\n```bash\npip install -e .\npip install --no-build-isolation flash-attn==2.4.2\n```\n\nIf you encounter out of RAM issue while installing [flash attention](https:\u002F\u002Fgithub.com\u002FDao-AILab\u002Fflash-attention?tab=readme-ov-file#installation-and-features), you can set the environment variable `MAX_JOBS` to limit the number of parallel compilation jobs:\n```bash\nMAX_JOBS=4 pip install --no-build-isolation flash-attn==2.4.2\n```\n\n### How to Get Started with Our AgiBot World Data \u003Ca name=\"startdata\">\u003C\u002Fa>\n\n#### Download Datasets \u003Ca name=\"downloaddatasets\">\u003C\u002Fa>\n\n- [OPTION 1] Download data from our [OpenDataLab](https:\u002F\u002Fopendatalab.com\u002FOpenDriveLab\u002FAgiBot-World) page.\n\n```bash\npip install openxlab # install CLI\nopenxlab dataset get --dataset-repo OpenDriveLab\u002FAgiBot-World # dataset download\n```\n\n- [OPTION 2] Download data from our [HuggingFace](https:\u002F\u002Fhuggingface.co\u002Fdatasets\u002Fagibot-world\u002FAgiBotWorld-Alpha) page.\n\n```bash\nhuggingface-cli download --resume-download --repo-type dataset agibot-world\u002FAgiBotWorld-Alpha --local-dir .\u002FAgiBotWorld-Alpha\n```\n\nConvert the data to **LeRobot Dataset** format following [any4lerobot](https:\u002F\u002Fgithub.com\u002FTavish9\u002Fany4lerobot).\n\n#### Visualize Datasets \u003Ca name=\"visualizedatasets\">\u003C\u002Fa>\n\nWe adapt and extend the dataset visualization script from [LeRobot Project](https:\u002F\u002Fgithub.com\u002Fhuggingface\u002Flerobot\u002Fblob\u002Fmain\u002Flerobot\u002Fscripts\u002Fvisualize_dataset.py):\n\n```bash\npython scripts\u002Fvisualize_dataset.py --task-id 390 --dataset-path \u002Fpath\u002Fto\u002Flerobot\u002Fformat\u002Fdataset\n```\n\nIt will open `rerun.io` and display the camera streams, robot states and actions, like this:\n\u003Cdiv style=\"text-align: center;\">\n\u003Cimg src=\"assets\u002Fdataset_visualization.gif\" width=\"600\">\n\u003C\u002Fdiv>\n\n### How to Get Started with Our GO-1 Model \u003Ca name=\"startmodel\">\u003C\u002Fa>\n\n#### Requirements \u003Ca name=\"requirements\">\u003C\u002Fa>\n\nWe strongly recommend full fine-tuning for the best performance. However, if GPU memory is limited, you can alternatively fine-tune only the Action Expert.\n\n|         Usage         |  GPU Memory Required  |     Example GPU     |\n| :-------------------: | :-------------------: | :-----------------: |\n|       Inference       |         ~7GB          |      RTX 4090       |\n|  Fine-tuning (Full)   | ~70GB (batch size=16) |   A100 80GB, H100   |\n| Fine-tuning (Only AE) | ~24GB (batch size=16) | RTX 4090, A100 40GB |\n\n#### Model Zoo \u003Ca name=\"modelzoo\">\u003C\u002Fa>\n\n|  Model   |                   HF Link                    |                              Description                              |\n| :------: | :------------------------------------------: | :-------------------------------------------------------------------: |\n| GO-1 Air | https:\u002F\u002Fhuggingface.co\u002Fagibot-world\u002FGO-1-Air | GO-1 model without Latent Planner pre-trained on AgiBot World dataset |\n|   GO-1   |   https:\u002F\u002Fhuggingface.co\u002Fagibot-world\u002FGO-1   |            GO-1 model pre-trained on AgiBot World dataset             |\n\n#### Fine-tuning on Your Own Dataset \u003Ca name=\"finetune\">\u003C\u002Fa>\n\nHere we provide an example of fine-tuning the GO-1 model on the [LIBERO](https:\u002F\u002Flibero-project.github.io\u002Fintro.html) dataset. You can easily adapt it for your own data.\n\n**1. Prepare Data**\n\nWe use the LeRobot dataset for our default dataset and dataloader. We provide a script for converting LIBERO to LeRobot format in [evaluate\u002Flibero\u002Fconvert_libero_data_to_lerobot.py](evaluate\u002Flibero\u002Fconvert_libero_data_to_lerobot.py).\n\nSince TensorFlow is required to read the [RLDS format](https:\u002F\u002Fgithub.com\u002Fgoogle-research\u002Frlds), we recommend creating a separate conda environment to avoid package conflicts:\n\n```bash\nconda create -n libero_data python=3.10 -y\nconda activate libero_data\n\npip install -e \".[libero_data]\"\n```\n\nDownload the raw LIBERO dataset from [OpenVLA](https:\u002F\u002Fhuggingface.co\u002Fdatasets\u002Fopenvla\u002Fmodified_libero_rlds), then run the script to convert it into LeRobot dataset:\n\n```bash\n# Optional: Change the LeRobot home directory\nexport HF_LEROBOT_HOME=\u002Fpath\u002Fto\u002Fyour\u002Flerobot\n\npython evaluate\u002Flibero\u002Fconvert_libero_data_to_lerobot.py --data_dir \u002Fpath\u002Fto\u002Fyour\u002Flibero\u002Fdata\n```\n\n**2. Prepare Configs**\n\nWe provide an example config for fine-tuning GO-1 on LIBERO in [go1\u002Fconfigs\u002Fgo1_sft_libero.py](go1\u002Fconfigs\u002Fgo1_sft_libero.py).\n\nKey sections in the config:\n- `DatasetArguments` - path or repo for the LeRobot dataset.\n- `GOModelArguments` - model settings: architecture (GO-1 Air or GO-1), action chunk size, diffusion scheduler, parameter freezing, etc.\n- `GOTrainingArguments` - training hyper-parameters, see [transformers docs](https:\u002F\u002Fhuggingface.co\u002Fdocs\u002Ftransformers\u002Fmain_classes\u002Ftrainer#transformers.TrainingArguments) for more details.\n- `SpaceArguments` - state\u002Faction dimensions, data keys in the LeRobot dataset, default language prompt, control frequency.\n\nSee [go1\u002Fconfigs\u002Fgo1_base_cfg.py](go1\u002Fconfigs\u002Fgo1_base_cfg.py) for all available config options.\n\n**3. Start Fine-tuning**\n\nStart fine-tuning with the following command, you can setup environment variables according to the [shell](go1\u002Fshell\u002Ftrain.sh).\n\n```bash\nRUNNAME=\u003CYOUR_RUNNAME> bash go1\u002Fshell\u002Ftrain.sh \u002Fpath\u002Fto\u002Fyour\u002Fconfig\n```\n\nCheckpoints will be saved in `experiment\u002F\u003CYOUR_RUNNAME>` and logs will be saved in `experiment\u002F\u003CYOUR_RUNNAME>\u002Flogs`.\n\n**Notes:**\n- We also provide a [debugging shell](go1\u002Fshell\u002Ftrain_dev.sh) which can run on a single RTX4090. It also set `DEBUG_MODE` to true for faster init. \n- We do not need to precompute the normalization statistics for the training data, as LeRobot will compute them when loading the dataset. The statistics will be saved to `experiment\u002F\u003CYOUR_RUNNAME>\u002Fdataset_stats.json`.\n- We set action chunk size and control frequency input as 30 in GO-1 pre-training, as our AgiBot World dataset is collected at 30Hz. We change them to 10 in LIBERO fine-tuning, as the LIBERO dataset is collected at 10Hz. You can change them accordingly in the config file.\n\n\n#### Testing Your Model \u003Ca name=\"inference\">\u003C\u002Fa>\n\n**Local Inference**\n\nAfter fine-tuning, you can test your model locally using an example script in [evaluate\u002Fdeploy.py](evaluate\u002Fdeploy.py). You can build a `GO1Infer` object to load the model and dataset statistics, then call the `inference` method to run inference:\n\n```python\nimport numpy as np\nfrom evaluate.deploy import GO1Infer\n\nmodel = GO1Infer(model_path=\"\u002Fpath\u002Fto\u002Fyour\u002Fcheckpoint\", data_stats_path=\"\u002Fpath\u002Fto\u002Fyour\u002Fdataset_stats.json\")\n\npayload = {\n    \"top\": ...,\n    \"right\": ...,\n    \"left\": ...,\n    \"instruction\": \"example instruction\",\n    \"state\": ...,\n    \"ctrl_freqs\": np.array([30]),\n}\n\nactions = model.inference(payload)\n```\n\nWe also provide a script for open-loop evaluation with training data in [evaluate\u002Fopenloop_eval.py](evaluate\u002Fopenloop_eval.py).\n\n**Remote Inference**\n\nConsidering that 1. real robot may not have powerful GPUs, 2. different robots and simulation benchmarks often require different package dependencies, we also provide a policy server for GO-1. A client in another environment or another machine send observations to the server for remote inference.\n\nStart the server and it will listen on port `PORT` and waits for observations:\n\n```bash\npython evaluate\u002Fdeploy.py --model_path \u002Fpath\u002Fto\u002Fyour\u002Fcheckpoint --data_stats_path \u002Fpath\u002Fto\u002Fyour\u002Fdataset_stats.json --port \u003CPORT>\n```\n\nFor the client, we provide a `GO1Client` class to send requests to the server and receive actions:\n\n```python\nfrom typing import Dict, Any\n\nimport json_numpy\nimport numpy as np\nimport requests\n\njson_numpy.patch()\n\nclass GO1Client:\n  def __init__(self, host: str, port: int):\n      self.host = host\n      self.port = port\n\n  def predict_action(self, payload: Dict[str, Any]) -> np.ndarray:\n      response = requests.post(\n          f\"http:\u002F\u002F{self.host}:{self.port}\u002Fact\", json=payload, headers={\"Content-Type\": \"application\u002Fjson\"}\n      )\n\n      if response.status_code == 200:\n          result = response.json()\n          action = np.array(result)\n          return action\n      else:\n          print(f\"Request failed, status code: {response.status_code}\")\n          print(f\"Error message: {response.text}\")\n          return None\n```\n\nWe can then run the LIBERO evaluation script to query the server, see the [LIBERO README](evaluate\u002Flibero\u002FREADME.md) for details.\n\n\n#### More Examples \u003Ca name=\"examples\">\u003C\u002Fa>\nWe will provide more examples of fine-tuning and running inference with GO-1 models on real robots and simulation platforms.\n\nCurrently we have:\n- [Genie Studio](https:\u002F\u002Fgenie.agibot.com\u002Fgeniestudio): AgiBot G1 with out-of-the-box GO-1 model plus integrated data collection, fine-tuning, and deployment pipeline.\n- [AgileX](evaluate\u002Fagilex\u002FREADME.md): AgileX Cobot Magic (Aloha)\n- [LIBERO](evaluate\u002Flibero\u002FREADME.md): LIBERO Simulation (Franka)\n- [RoboTwin](https:\u002F\u002Fgithub.com\u002FRoboTwin-Platform\u002FRoboTwin\u002Ftree\u002Fmain\u002Fpolicy\u002FGO1): RoboTwin Simulation (Aloha)\n\n\u003C!-- \u003Cp align=\"right\">(\u003Ca href=\"#top\">back to top\u003C\u002Fa>)\u003C\u002Fp> -->\n\n\n\n\u003C!-- \u003Cp align=\"right\">(\u003Ca href=\"#top\">back to top\u003C\u002Fa>)\u003C\u002Fp> -->\n\n\n\n## 📄 License and Citation   \u003Ca name=\"liscenseandcitation\">\u003C\u002Fa>\n\nAll the data and code within this repo are under [CC BY-NC-SA 4.0](https:\u002F\u002Fcreativecommons.org\u002Flicenses\u002Fby-nc-sa\u002F4.0\u002F). \n\n- Please consider citing our work if it helps your research.\n- For the full authorship and detailed contributions, please refer to [contributions](CONTRIBUTING.md).\n- In alphabetical order by surname:\n```BibTeX\n@article{bu2025agibot_arxiv,\n  title={Agibot world colosseo: A large-scale manipulation platform for scalable and intelligent embodied systems},\n  author={Bu, Qingwen and Cai, Jisong and Chen, Li and Cui, Xiuqi and Ding, Yan and Feng, Siyuan and Gao, Shenyuan and He, Xindong and Huang, Xu and Jiang, Shu and others},\n  journal={arXiv preprint arXiv:2503.06669},\n  year={2025}\n}\n\n@inproceedings{bu2025agibot_iros,\n  title={Agibot world colosseo: A large-scale manipulation platform for scalable and intelligent embodied systems},\n  author={Bu, Qingwen and Cai, Jisong and Chen, Li and Cui, Xiuqi and Ding, Yan and Feng, Siyuan and He, Xindong and Huang, Xu and others},\n  booktitle={2025 IEEE\u002FRSJ International Conference on Intelligent Robots and Systems (IROS)},\n  year={2025},\n  organization={IEEE}\n}\n\n@article{shi2025diversity,\n  title={Is Diversity All You Need for Scalable Robotic Manipulation?},\n  author={Shi, Modi and Chen, Li and Chen, Jin and Lu, Yuxiang and Liu, Chiming and Ren, Guanghui and Luo, Ping and Huang, Di and Yao, Maoqing and Li, Hongyang},\n  journal={arXiv preprint arXiv:2507.06219},\n  year={2025}\n}\n```\n\n## 📝 Blogs  \u003Ca name=\"blogs\">\u003C\u002Fa>\n```BibTeX\n@misc{AgiBotWorldTeam2025agibot-world-colosseo,\n          title        = {Introducing AgiBot World Colosseo: A Large-scale Manipulation Platform for Scalable and Intelligent Embodied Systems},\n          author       = {Shi, Modi and Lu, Yuxiang and Wang, Huijie and Xie, Chengen and Bu, Qingwen},\n          year         = {2025},\n          month        = {March},\n          howpublished = {\\url{https:\u002F\u002Fopendrivelab.com\u002FAgiBot-World\u002F}},\n          note         = {Blog post},\n        }\n\n@misc{AgiBotWorldTeam2025open-sourcing-go1,\n          title        = {Open-sourcing GO-1: The Bitter Lessons of Building VLA Systems at Scale},\n          author       = {Shi, Modi and Lu, Yuxiang and Wang, Huijie and Yang, Shaoze},\n          year         = {2025},\n          month        = {September},\n          howpublished = {\\url{https:\u002F\u002Fopendrivelab.com\u002FOpenGO1\u002F}},\n          note         = {Blog post},\n        }\n```\n","AgiBot-World是一个用于推进大规模智能实体系统双臂操作的全栈机器人学习平台。该项目提供了基础模型、基准测试和生态系统，以促进学术界和工业界对高质量机器人数据的访问，旨在为具身AI带来“ImageNet时刻”。其核心功能包括预训练的机器人基础模型GO-1及其轻量级版本GO-1 Air，以及包含超过一百万条轨迹的大规模数据集AgiBot World Beta。此外，项目还提供了一个任务目录，详细描述了数据集中涉及的任务类型和样本动作文本。适合于需要开发或研究高级机器人操作技能的应用场景，如自动化制造、服务机器人等。",2,"2026-06-11 03:41:32","high_star"]