3D Logic on Super-CLEVR
97Pass@1MIG
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| MIGevaluation_mode=In-Domain, training_protocol=Task-Specific Training2026.02 | 97 | 99 | 8 | |
| GRPOevaluation_mode=In-Domain, training_protocol=Task-Specific Training2026.02 | 89 | 97 | — | |
| Base Modelevaluation_mode=In-Domain, training_protocol=Task-Specific Training2026.02 | 46 | 80 | — |