| Task Name | Dataset Name | SOTA Result | Trend | |
|---|---|---|---|---|
| Context window selection | DiffReact PDEBench (full-oracle case-level) | Lsel5 | 24 | |
| Regret assessment | DiffReact full-oracle case-level PDEBench | Regret Knee-0.8 | 16 | |
| Trajectory Rollout Prediction | DiffReact f2 dense (OOD) | OOD Mean Error0.086 | 5 |