Arithmetic Reasoning on MultiArith baseline original
593Correct CountGPT-OSS
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| GPT-OSSParameters=120B, N=6002026.06 | 593 | 98.83 | 27.39 | |
| Gemma4Parameters=31B, N=6002026.06 | 587 | 97.83 | 120.62 | |
| DeepSeek-R1Parameters=70B, N=6002026.06 | 574 | 95.67 | 555.49 |