Mathematical Reasoning on MATH (Normalized Spearman distance)
13Normalized Spearman DistanceTrunc50→40
Evaluation results
| Method | Links | |
|---|---|---|
| Trunc50→40Perturbation Level=Task-prompt2026.06 | 13 | |
| SingleGraph1Perturbation Level=Task-prompt2026.06 | 18 | |
| ScoreWin80Perturbation Level=Task-prompt2026.06 | 20 | |
| NoTrunc bootPerturbation Level=Task-prompt2026.06 | 21 | |
| Block top-2Perturbation Level=Task-prompt2026.06 | 23 | |
| Random50→40×10Perturbation Level=Task-prompt2026.06 | 24 | |
| Block top-1Perturbation Level=Task-prompt2026.06 | 31 | |
| Trunc25→20Perturbation Level=Judge-prompt with Prometheus2026.06 | 31 | |
| Block top-2Perturbation Level=Judge-prompt with Prometheus2026.06 | 32 | |
| ScoreWin40Perturbation Level=Judge-prompt with Prometheus2026.06 | 37 | |
| Block top-1Perturbation Level=Judge-prompt with Prometheus2026.06 | 38 | |
| NoTrunc bootPerturbation Level=Judge-prompt with Prometheus2026.06 | 43 | |
| Random50→20×10Perturbation Level=Judge-prompt with Prometheus2026.06 | 44 | |
| SingleGraph1/all10Perturbation Level=Judge-prompt with Prometheus2026.06 | 44 |