List your company

Continuous Control (MuJoCo) on Hopper v4

3,240Mean Episodic ReturnFP16-PPO
2,740.82,870.43,0003,129.6Apr 27, 2026

Evaluation results

MethodLinks
FP16-PPOPrecision=FP162026.043,240—
INT8-PPOPrecision=INT82026.043,180—
INT4-PPOPrecision=INT42026.043,050—
BitRL-1bitPrecision=1-bit2026.042,89089.2
MLP-PPOArchitecture=MLP2026.042,760—