← Leaderboard

H-RDT

Policy
Tsinghua University & Horizon Robotics · Jul 31, 2025

Bimanual diffusion-transformer policy pretrained on egocentric human manipulation data before robot fine-tuning. Results are self-reported (58% average on two real bimanual tasks vs 34% for RDT); its RoboTwin 2.0 results cover task subsets only.

Type
Policy
Parameters
2B
Backbone
Diffusion transformer pretrained on EgoDex human video
Weights
Open
License
MPL-2.0
Commercial use
Unclear
Released
Jul 31, 2025
Robots
AgileX ALOHA-2, Dual UR5 (UMI), Aloha-AgileX (sim)
SO-101
–
LeRobot
–
Inputs
–
VRAM
–

Benchmark results

1 of 13 benchmarks
#51 of 540.1%
Not reported on 12 benchmarks
Wrong or missing number? Report an error