Mathematical Reasoning on MultiArith OOD
100Base Accuracy (CA)THOUGHTSTEER
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| THOUGHTSTEERArchitecture=SimCoT (Llama-3.2) 3B, Decoding method=CODI2026.04 | 100 | 100 | 100 | |
| THOUGHTSTEERArchitecture=SimCoT (Llama-3.2) 1B, Decoding method=CODI2026.04 | 97.8 | 97.8 | 98.9 |