Mathematical Reasoning on SVAMP OOD
73Base AccuracyTHOUGHTSTEER
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| THOUGHTSTEERArchitecture=SimCoT (Llama-3.2) 3B, Decoding method=CODI2026.04 | 73 | 73 | 99.7 | |
| THOUGHTSTEERArchitecture=SimCoT (Llama-3.2) 1B, Decoding method=CODI2026.04 | 61.5 | 62 | 94.4 |