Language-Conditioned Tasks on SmartHome Light
0.85Mean Episodic ReturnFP16-PPO
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| FP16-PPOPrecision=FP162026.04 | 0.85 | — | |
| INT8-PPOPrecision=INT82026.04 | 0.84 | — | |
| INT4-PPOPrecision=INT42026.04 | 0.82 | — | |
| BitRL-1bitPrecision=1-bit2026.04 | 0.82 | 96.5 | |
| MLP-PPOArchitecture=MLP2026.04 | 0.71 | — |