InternW0-Libero

Model description

InternW0-Libero is the task-posttrained InternW0 checkpoint for LIBERO. It is derived from the heterogeneous InternW0-Base pretraining checkpoint and posttrained with the R13 Goal+Long recipe. The selected step-600 checkpoint is evaluated on all four official LIBERO suites.

InternW0 couples a slow Wan2.2-TI2V-5B VideoDiT planner with a fast ActionDiT executor. Observation-conditioned chunk K/V routing adapts cached future-video context to newly observed states, so the model can reuse long-horizon predictions while generating responsive action chunks.

This is a native w0 checkpoint, not a standalone Transformers model. It is intended for LIBERO evaluation or compatible continued posttraining.

Files and integrity

pytorch_model.bin is a second Git-LFS path to the exact same object as InternW0-Libero.pt. It does not duplicate the 14.7 GB payload.

Checkpoint summary

• Selected step: 600
• Serialization dtype: bfloat16
• Model-only checkpoint: yes
• Optimizer state: excluded
• Tensors: 1,977
• Parameter elements: 7,352,528,643
• Source heterogeneous pretraining step: 30000
• Target domain: libero/franka
• Target soft prompts: K=8
• Canonical action/proprio dimensionality: 37
• Native EEF10 slots: 10
• Action horizon: 64
• Action chunk size: 16
• Video plan horizon: 96 action ticks
• Video/action stride: 8
• Maximum action offset: 15
• History frames: 0
• Action semantics: native delta EEF10, row-aligned

The checkpoint embeds a schema-v4 behavior_config, training_recipe, posttrain_statistics, and posttrain_provenance. The public loader performs a semantic contract check before loading weights.

LIBERO training

The R13 recipe uses:

• posttraining suites libero_goal and libero_10;
• random-frame sampling with 100,000 samples per epoch;
• 10 epochs;
• batch size 48 per process;
• learning rate 5e-5 with cosine decay;
• seed 42;
• native 20 Hz EEF10 actions and proprio;
• action offset sampled in [0, 15];
• action horizon 64 with four 16-action chunks;
• K=8 target soft prompts.

The matching normalization statistics are in normalization.json. The R13 run initialized from a preceding LIBERO checkpoint at step 2,640 and selected step 600 over the final step 660 based on the complete four-suite evaluation. See PROVENANCE.md for the full lineage.

Evaluation result

The selected checkpoint was evaluated on all 40 tasks in the four official LIBERO suites, with 50 trials per task:

The complete per-task record is included in evaluation_summary.json. The evaluation uses environment/init seed 42, an unset policy diffusion seed, 30 settling steps, execution of all 16 actions per chunk, 10 inference steps, runtime gripper binarization, lossless PNG transport, 400-step limits for Spatial/Object/Goal, a 700-step limit for libero_10, and MuJoCo 3.3.2.

Required external assets

This checkpoint is not self-contained. Users must separately obtain and comply with the licenses for:

hf download Wan-AI/Wan2.2-TI2V-5B --local-dir assets/wan22
hf download facebook/dinov3-vitb16-pretrain-lvd1689m \
  --local-dir assets/dinov3-vitb16-pretrain-lvd1689m

A LIBERO checkout, the matching LIBERO LeRobot conversion, and its UMT5 text cache are also required. The repository does not bundle datasets, simulation code, Wan2.2 files, or DINOv3 weights.

Quick start

1. Install the public codebase

git clone https://github.com/InternRobotics/InternW0.git
cd InternW0

conda create -n w0 python=3.10 -y
conda activate w0
pip install -U pip setuptools wheel
pip install torch==2.7.1 torchvision==0.22.1 \
  --extra-index-url https://download.pytorch.org/whl/cu128
pip install -e .

2. Download this checkpoint

hf download InternRobotics/InternW0-Libero \
  config.json pytorch_model.bin checkpoint_metadata.json \
  normalization.json posttrain_config.yaml evaluation_config.yaml \
  source_training_config.yaml source_evaluation_config.yaml \
  evaluation_summary.json evaluation_meta.json PROVENANCE.md LICENSE \
  --local-dir assets/libero

mv assets/libero/pytorch_model.bin \
   assets/libero/InternW0-Libero.pt

3. Run LIBERO evaluation

export CKPT="$PWD/assets/libero/InternW0-Libero.pt"
export W0_WAN_MODEL_DIR="$PWD/assets/wan22"
export W0_DINO_WEIGHTS="$PWD/assets/dinov3-vitb16-pretrain-lvd1689m"
export W0_LIBERO_STATS="$PWD/assets/libero/normalization.json"
export W0_LIBERO_TEXT_CACHE="$PWD/assets/libero/text_embeds_cache"
export LIBERO_PATH="$PWD/external/LIBERO"

SUITE=libero_goal TASK_ID=0 NUM_TRIALS=50 \
bash scripts/run_libero_eval.sh

For compatible continued posttraining:

bash scripts/train_zero1.sh 8 --config-name posttrain_libero

Download statistics

Hugging Face uses library-specific query files for model download statistics. For the pytorch library, the Hub counts HTTP GET and HEAD requests to pytorch_model.bin. This repository provides that path as a zero-copy alias of the canonical checkpoint. The badge above reads the public Hub API downloads field, corresponding to downloads in the last 30 days. Hub analytics are aggregated and may lag briefly.

Intended uses

• LIBERO reproduction and extension.
• Research on world-action models, context routing, and cross-embodiment robot learning.
• Compatible continued posttraining with the public w0 codebase.

Limitations

• This is a simulation-domain research checkpoint, not a deployed robot controller.
• It requires the public InternW0 codebase, external Wan2.2 and DINOv3 assets, LIBERO data, text caches, and substantial compute.
• It is not directly compatible with transformers.AutoModel.
• Real-robot use requires independent safety analysis, constrained workspaces, and human oversight.

License

The weights and metadata in this repository are released under the MIT License. Third-party assets obtained separately remain subject to their original licenses.

Citation

If you use InternW0, please cite:

@article{internw0,
  title = {InternW0: A Foundational Physical World Model for Efficient Real-World Interactions},
  author = {Jisong Cai and Yao Mu and Ganlin Yang and Zhe Cao and Zhangzheng Tu and Xing Gao and Kailin Li and Xinyu Zhan and Lixin Yang and Yangkun Zhu and Haoxiang Ma and Ming Zhou and Qiaojun Yu and Yufei Xue and Liqun He and Yifei Yao and Yifan Zhu and Long Ling and Bingqi Jiang and Haoyu Guo and Xueyue Zhu and Bowen Zhou and Bin Zhao and Tianfan Xue and Chunhua Shen and Weinan Zhang},
  journal = {arXiv preprint arXiv:2609.27656},
  year = {2026}
}