Meituan-Robotics-0-RoboTwin
MR0 policy adapted to the RoboTwin benchmark. The selected model achieves 93.22% mean success on the clean setting and 93.88% on the randomized setting across 50 tasks.
Code · Getting started · File checksums
| Property | Value |
|---|---|
| Framework | QwenMMDiT |
| Action dimensions | 14 |
| Action horizon | 100 |
| Weights | checkpoints/model.pt |
| Configuration | config.full.yaml |
| Normalization statistics | dataset_statistics.json |
Download and load
Install the MR0 code at commit 88c66793 or later, following the Getting Started guide. This loader reads the complete VLA checkpoint without a separate Qwen weight file. Then download this repository:
import os
from pathlib import Path
from huggingface_hub import snapshot_download
root = Path(snapshot_download(
repo_id="MeituanRobotics/Meituan-Robotics-0-RoboTwin",
local_dir="Meituan-Robotics-0-RoboTwin",
)).resolve()
os.environ["MR0_RELEASE_DIR"] = str(root)
from mr0.model.framework.base_framework import baseframework
model = baseframework.from_pretrained(root / "checkpoints/model.pt")
model = model.eval().to("cuda")
checkpoints/model.pt contains the complete Qwen3.5 backbone and MMDiT action head. base_vlm/Qwen3.5-4B-Action/ contains only the matching tokenizer, processor, and architecture configuration; it is not a second model. The release also includes the VLA configuration and normalization statistics. Use new_embodiment as the normalization key and retain the included image-processing settings.
The checkpoint bytes are unchanged. release.json records the model provenance and SHA256SUMS records file identities. For downstream training, adapt the dataset settings and set both framework.qwenvl.checkpoint_path and trainer.pretrained_checkpoint to the downloaded complete checkpoint. Set framework.qwenvl.base_vlm to its metadata directory.
File verification
SHA256SUMS is the single SHA-256 checksum manifest for this release. It covers every repository file except the manifest itself, including weights, configuration, tokenizer/processor metadata, and documentation.
After downloading the complete repository, run from its root directory:
sha256sum -c SHA256SUMS