Logical Reasoning on ZebraLogic (held-out)
71.7AccuracyCORE
Evaluation Results
| Method | Links | |
|---|---|---|
| COREtraining samples=102026.05 | 71.7 | |
| GEPAtraining samples=1002026.05 | 70.7 | |
| COREtraining samples=52026.05 | 70 | |
| MemRLtraining samples=52026.05 | 68.3 | |
| COREtraining samples=1002026.05 | 66.3 | |
| GEPAtraining samples=52026.05 | 59.7 | |
| MemRLtraining samples=1002026.05 | 58.7 | |
| GEPAtraining samples=102026.05 | 57 | |
| MemRLtraining samples=102026.05 | 54.3 | |
| Episodic RAGtraining samples=52026.05 | 54 | |
| GRPOtraining samples=102026.05 | 53.3 | |
| GRPOtraining samples=52026.05 | 52.3 | |
| GRPOtraining samples=1002026.05 | 52 | |
| No Learningtraining samples=52026.05 | 50.9 | |
| No Learningtraining samples=102026.05 | 50.9 | |
| No Learningtraining samples=1002026.05 | 50.9 | |
| Episodic RAGtraining samples=1002026.05 | 49.7 | |
| Episodic RAGtraining samples=102026.05 | 49.3 |