Alignment ranking evaluation on GDPVal 44 tasks
0.8722Mean NDCG@8GSPO Rubric
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| GSPO RubricModel–Rubric Pair=GSPO Rubric2025.12 | 0.8722 | 0.0985 | |
| SFT RubricModel–Rubric Pair=SFT Rubric2025.12 | 0.8103 | 0.0881 | |
| No-Conv (Base)Model–Rubric Pair=No-Conv (Base)2025.12 | 0.7998 | 0.0807 |