RobustHumanoid PPO Results: Policy Performance Grid for Humanoid-v4

RobustHumanoid benchmark results evaluate a Stable-Baselines3 PPO policy on the Humanoid-v4 environment under out-of-distribution MuJoCo physics. The policy was tested over a 20 x 20 grid of ground friction and actuator torque scale values, with 5 episodes per grid point. Author kaus4004 uploaded the dataset to Hugging Face in July 2026.

Use Cases

• Analyze policy robustness based on systematic variation of physics parameters
• Benchmark reinforcement learning algorithms based on out-of-distribution performance
• Study the sensitivity of humanoid control policies based on ground friction and actuator torque scales

Strengths

• Systematic evaluation over a 20 x 20 parameter grid
• 5 episodes per grid point for statistical reliability
• Focus on out-of-distribution robustness testing

Limitations

• Column-level documentation is absent; field semantics must be inferred after download
• Row count is unknown, which may limit suitability assessment

Provenance

Source

huggingface

Collection Method

Benchmark evaluation of a trained PPO policy

Freshness

Last updated 2026-07-22 13:49:32