Speaking Head Motion Generation on Seamless Interaction Dataset
2.95LVEGDPO-Listener
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| GDPO-Listener2026.03 | 2.95 | 1.06 | 22.45 | 2.71 | 1.26 | |
| L2L2026.03 | 3.1 | 1.25 | 53.03 | 3.52 | 3.32 | |
| DualTalk2026.03 | 6.53 | 1.95 | 37.46 | 9.7 | 1.02 | |
| ARTalk*adapted for speak-listen generation=true, additional partner audio input=true2026.03 | 6.79 | 2.02 | 27.41 | 8.55 | 0.81 | |
| ARTalk2026.03 | 7.46 | 2.12 | 31.64 | 7.66 | 1.19 | |
| DiffposeTalk2026.03 | 9.48 | 2.96 | 32.66 | 7.89 | 1.4 |