Arithmetic on arithmetic LLAMA-3.2-3B-INSTRUCT (test)
70.2Avg@16RLVR
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| RLVRModel=LLAMA-3.2-3B-INSTRUCT, Mode=Zero-shot2025.10 | 70.2 | 87.8 | |
| RLERModel=LLAMA-3.2-3B-INSTRUCT, Mode=Zero-shot, k=4, Gk=82025.10 | 63.9 | 75.2 | |
| RLERModel=LLAMA-3.2-3B-INSTRUCT, Mode=Zero-shot, k=4, Gk=42025.10 | 62.3 | 74.4 | |
| RLERModel=LLAMA-3.2-3B-INSTRUCT, Mode=Zero-shot, k=8, Gk=22025.10 | 61.9 | 78.6 | |
| RLERModel=LLAMA-3.2-3B-INSTRUCT, Mode=Zero-shot, k=2, Gk=82025.10 | 58.7 | 69.8 | |
| RLIR INTUITORModel=LLAMA-3.2-3B-INSTRUCT, Mode=Zero-shot2025.10 | 50.3 | 54.2 | |
| RLIR SCModel=LLAMA-3.2-3B-INSTRUCT, Mode=Zero-shot2025.10 | 48.4 | 54.4 | |
| RLIR FreqModel=LLAMA-3.2-3B-INSTRUCT, Mode=Zero-shot2025.10 | 47.5 | 56.2 | |
| RLIR JudgeModel=LLAMA-3.2-3B-INSTRUCT, Mode=Zero-shot2025.10 | 41.7 | 77.3 | |
| pre-RLModel=LLAMA-3.2-3B-INSTRUCT, Mode=Zero-shot2025.10 | 30.7 | 80.2 |