molmospaces

# MolmoSpaces This respository contains asset data for [MolmoSpaces](https://github.com/allenai/molmospaces), including - Objects - Robots - Scenes - Grasps - Benchmarks ## Updates - **[2026/05/28] - New mujoco scene versions (`ithor`, `procthor-10k-{train,val,test}`, `procthor-objaverse-{train,val}`, and `holodeck-objaverse-{train,val}`) with included occupany maps (suffix `_with_occupancy`)** - **[2026/02/16] - Isaac-compatible USD objects and scenes now also available** ## Downloading We recommend using the [download.py](download.py) script to list available data sources and extract data from one or all sources to a local cache directory. To use the script you'll need a few dependecies in your Python environment that can be installed, e.g., by: ```bash pip install zstandard datasets huggingface-hub tqdm ``` ## License The Objaverse subset is licensed under [ODC-BY 1.0](https://opendatacommons.org/licenses/by/1-0/). All other data subsets are licensed under [CC BY 4.0](https://creativecommons.org/licenses/by/4.0/legalcode.en). This dataset is intended for research and educational use in accordance with Ai2's [Responsible Use Guidelines](https://allenai.org/responsible-use). The scripts are licensed under [Apache 2.0](CODE_LICENSE). ## Data Attributions The xml files have been modified from the original versions provided by the following sources: - [mujoco_menagerie / franka_fr3](https://github.com/google-deepmind/mujoco_menagerie/tree/main/franka_fr3) - Developed by Franka Robotics - [mujoco_menagerie / robotiq_2f85_v4](https://github.com/google-deepmind/mujoco_menagerie/tree/main/robotiq_2f85_v4) - Copyright (c) 2013, ROS-Industrial - [Rainbow Robotics / rby1-sdk](https://github.com/RainbowRobotics/rby1-sdk) - Copyright 2024-2025 Rainbow Robotics - [RUM Gripper](https://github.com/jeffacce/cap-policy) - Copyright (c) 2026 NYU Generalizable Robotics and AI Lab (GRAIL) The included metadata (downloadable from the `mujoco/objects/objathor_metadata/20260129` source) provides the license for each object in the Objaverse subset. We note that Objaverse models have been significantly modified to reduce memory and processing requirements, including mesh decimation, convex collider extraction, and baking of visual effects via Blender scripts, so the provided quality may not reflect the original model. ## BibTeX ``` @misc{molmospaces2026, title={MolmoSpaces: A Large-Scale Open Ecosystem for Robot Navigation and Manipulation}, author={Yejin Kim and Wilbert Pumacay and Omar Rayyan and Max Argus and Winson Han and Eli VanderBilt and Jordi Salvador and Abhay Deshpande and Rose Hendrix and Snehal Jauhri and Shuo Liu and Nur Muhammad Mahi Shafiullah and Maya Guru and Arjun Guru and Ainaz Eftekhar and Karen Farley and Donovan Clay and Jiafei Duan and Piper Wolters and Alvaro Herrasti and Ying-Chun Lee and Georgia Chalvatzaki and Yuchen Cui and Ali Farhadi and Dieter Fox and Ranjay Krishna}, year={2026}, eprint={2602.11337}, archivePrefix={arXiv}, primaryClass={cs.RO}, url={https://arxiv.org/abs/2602.11337}, } ```