Mathematical Reasoning on MATH500 (Acc., Steps, TPF)
40.6AccuracyKLASS
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| KLASSBase Decoder=Dream2026.07 | 40.6 | 148.85 | 1.72 | |
| KLASS + TACGBase Decoder=Dream2026.07 | 40.6 | 145.79 | 1.76 | |
| Conf. + TACGBase Decoder=Dream2026.07 | 40 | 80.33 | 3.19 | |
| Conf.Base Decoder=Dream2026.07 | 39 | 94.6 | 2.71 | |
| d3LLM†Base Decoder=Dream2026.07 | 38.2 | — | 3.92 |