Sim-to-real adaptation on DoTA (target)
65.5AccuracyCURVE
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| CURVESupervision level=Full (Oracle)2026.01 | 65.5 | 69.42 | 30.33 | |
| RS2GSupervision level=Full (Oracle), Confidence threshold=0.62026.01 | 65.08 | 70.16 | 29.16 | |
| RuleSupervision level=Full (Oracle)2026.01 | 64.43 | 69.6 | 29.16 | |
| RS2GSupervision level=Full (Oracle), Confidence threshold=0.252026.01 | 64.35 | 69.46 | 28.66 | |
| CURVESupervision level=10%2026.01 | 64.35 | 69.02 | 27.91 | |
| CURVESupervision level=5%2026.01 | 63.72 | 67.94 | 27.47 | |
| RS2GSupervision level=10%, Confidence threshold=0.62026.01 | 62.68 | 65.12 | 24.17 | |
| RuleSupervision level=10%2026.01 | 59.45 | 63.36 | 17.24 | |
| RS2GSupervision level=10%, Confidence threshold=0.252026.01 | 59.17 | 61.38 | 16.78 | |
| CURVESupervision level=1%2026.01 | 58.8 | 60.58 | 15.65 | |
| CURVESupervision level=Zero-shot2026.01 | 58.5 | 59.26 | 15.55 | |
| RS2GSupervision level=1%, Confidence threshold=0.252026.01 | 58.28 | 59.79 | 15.5 | |
| RS2GSupervision level=5%, Confidence threshold=0.252026.01 | 58.17 | 62.42 | 15.11 | |
| RuleSupervision level=5%2026.01 | 57.87 | 60.17 | 13.68 | |
| RS2GSupervision level=5%, Confidence threshold=0.62026.01 | 57.24 | 59.53 | 12.28 | |
| RS2GSupervision level=Zero-shot, Confidence threshold=0.252026.01 | 56.76 | 57.36 | 13.11 | |
| RS2GSupervision level=1%, Confidence threshold=0.62026.01 | 56.76 | 59.26 | 12.07 | |
| RuleSupervision level=1%2026.01 | 55.98 | 58.64 | 12.43 | |
| RS2GSupervision level=Zero-shot, Confidence threshold=0.62026.01 | 55.98 | 57 | 11.48 | |
| RuleSupervision level=Zero-shot2026.01 | 55.07 | 57.35 | 6.3 |