| Benchmark | Model | Rank | Results |
|---|---|---|---|
| instruction-following-on-ifbench | TGOPD (4B multi-domain GRPO-fallback) | – | Accuracy: 51.72 |
| mathematical-reasoning-on-aime-2025 | TGOPD (35B multi-domain mask-only) | – | Accuracy: 74.22 |
| mathematical-reasoning-on-aime-2026 | TGOPD (35B single-domain GRPO-fallback) | – | Accuracy: 78.7 |
| mathematical-reasoning-on-hmmt-feb-2025 | TGOPD (35B single-domain mask-only) | – | Accuracy: 64.27 |