Inter-model agreement on STRUCTDEBATE (Pool (P))
0.623Kendall TauGRASP
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| GRASPnormalization=none, rescale defense matrix=false2026.05 | 0.623 | 0.189 | 0.78 | 50.9 | 57.4 | |
| GRASP-W∞normalization=L∞, rescale defense matrix=false2026.05 | 0.623 | 0.188 | 0.781 | 50.9 | 57.4 | |
| GRASP-W1normalization=L1, rescale defense matrix=false2026.05 | 0.604 | 0.198 | 0.772 | 52.8 | 58.7 | |
| GRASP-W∞+D¯normalization=L∞, rescale defense matrix=true2026.05 | 0.591 | 0.205 | 0.761 | 53 | 58.4 | |
| GRASP-W1+D¯normalization=L1, rescale defense matrix=true2026.05 | 0.58 | 0.21 | 0.752 | 53.5 | 59.3 | |
| RAW+SSstructural sufficiency guidance=true2026.05 | 0.393 | 0.303 | 0.405 | 20.8 | 23.5 | |
| RAW2026.05 | 0.337 | 0.331 | 0.425 | 38.5 | 41.7 |