Arithmetic Reasoning on MultiArith numeric-remapping (attack set)
100AccuracyGPT-OSS
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| GPT-OSSParameters=120B2026.06 | 100 | 0 | 5.78 | |
| Gemma4Parameters=31B2026.06 | 98.42 | 1.58 | 24.03 | |
| DeepSeek-R1Parameters=70B2026.06 | 98.25 | 1.75 | 111.46 |