WB-WAM

HumanoidArena

中文说明

This repository contains seven HumanoidArena post-train checkpoints under humanoid_arena_native50/: box_shelf, hammer, kick_football, obstacle_navigation, open_door, punch_markers, and sit_sofa. Each task directory has a checkpoint, its resolved training config.yaml, and dataset_stats.json.

Post-training uses official SONIC reference actions at their native 50 Hz rate, with separate normalization statistics for each task.

Download the HumanoidArena checkpoints into a local directory:

hf download WB-WAM/WB-WAM-HumanoidArena --local-dir ./checkpoints/wbwam

Training and evaluation instructions are in the WB-WAM repository.

Evaluation results

These results were obtained with Isaac Sim 5.1.0 on an NVIDIA RTX 5090 GPU, using three seeds and 20 episodes per seed for each task. Values are the mean ± population standard deviation of success rates across seeds.

Task Success rate
Football 70.0 ± 8.2%
SitSofa 95.0 ± 4.1%
Vision Navigation 76.7 ± 4.7%
DoubleDesk 65.0 ± 4.1%
P&PBox 86.7 ± 2.4%
Boxing 81.7 ± 2.4%
OpenDoor 98.3 ± 2.4%
Overall (21 task-seed success rates) 81.9 ± 12.3%
Downloads last month

-

Downloads are not tracked for this model. How to track
Video Preview
loading