RoboTwin 2.0, success rate (%) over 50 tasks, as reported in the paper (Table 1).
Clean and Randomized are background conditions; the two rows are inference
regimes served by these same weights.
Inference regime
Clean
Randomized
Avg.
Action-only
94.5
94.6
94.6
Full joint
94.3
94.8
94.6
K = 4 Euler denoising steps throughout. Full joint additionally denoises the
future-video, DINO, and pointmap streams; action-only skips them, trading them
for cheaper inference at the same average success rate. Selecting a regime is an
inference-time flag — no retraining, no separate weights.
Architecture
A Mixture-of-Transformers pairing a video DiT with an action DiT, coupled by
HBridge. Alongside actions the model can denoise three auxiliary streams:
future video, DINO features, and pointmaps.
14-d each (bimanual ALOHA-AgileX), ConcatLeftAlign
Action : video rate
4:1
Training
Data
2,500 clean + 25,000 randomized demos, all 50 tasks
Epochs
6
Learning rate
1e-4
Precision
bf16
Trained with flex-joint sampling at p = 0.5 on every present and joint flag,
with cross-modal prediction enabled for all three streams. That is what lets one
set of weights serve any regime in the results table above.
Files
text
1config.yaml # architecture + processor; autoloaded by the eval
2dataset_stats.json # action/state normalization statistics
3checkpoints/weights/step_048060.pt # 12 GB
Keep this directory layout. The eval locates config.yaml and
dataset_stats.json by walking up from the checkpoint path.
Usage
This repository holds the policy weights only. The Wan2.2 base components
and the ActionDiT backbone are separate downloads, resolved through
DIFFSYNTH_MODEL_BASE_PATH — see docs/INSTALL.md
and docs/ROBOTWIN.md.
The launcher defaults reproduce the full joint row: NUM_INFERENCE_STEPS=4,
INSTRUCTION_TYPE=unseen, EVAL_NUM_EPISODES=100, and all six regime flags
true. For the action-only row, set the three INFER_JOINT_* flags to
false. PHASES=clean,random covers both background conditions.
1@article{yan2026flexpi,
2 title = {Flex-$\pi$: A Multi-Stream World-Action Model with Compute Flexibility},
3 author = {Yan, Ge and Liu, Jinghao and Fan, Yuzhi and Cai, Lei and Liao, Minwen
4 and Zhang, Jesse and Fox, Dieter},
5 journal = {arXiv preprint arXiv:2608.10860},
6 year = {2026},
7 url = {https://arxiv.org/abs/2608.10860}
8}
