Multi-task Reinforcement Learning on Meta-World MT50
81Success Rate (d=0)ACNet
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| ACNetBackbone=Evo-1, Latency (ms)=91 ms, Frequency (Hz)=11.0 Hz2026.06 | 81 | 76 | 74 | 73 | 74 | |
| Naïve AsyncLatency (ms)=73 ms, Frequency (Hz)=13.6 Hz2026.06 | 80 | 71 | 70 | 70 | 70 | |
| Training-RTCBackbone=π0, Latency (ms)=134 ms, Frequency (Hz)=7.46 Hz2026.06 | 80 | 77 | 74 | 73 | 74 | |
| RTCBackbone=π0, Latency (ms)=159 ms, Frequency (Hz)=6.28 Hz2026.06 | 79 | 72 | 72 | 71 | 71 |