Mathematical Reasoning on GSM8K (Accuracy, Steps, TPF)
79.68AccuracyKLASS
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| KLASSBase Decoder=Dream2026.07 | 79.68 | 155.96 | 1.64 | |
| KLASS + TACGBase Decoder=Dream2026.07 | 78.62 | 91.94 | 2.78 | |
| d3LLM†Base Decoder=Dream2026.07 | 77.2 | — | 4.8 | |
| Conf. + TACGBase Decoder=Dream2026.07 | 73.31 | 69.01 | 3.71 | |
| Conf.Base Decoder=Dream2026.07 | 73.16 | 74.63 | 3.43 |