Mathematical Reasoning on Countdown, GSM8K, MATH500, and SVAMP
54.92AccuracyDiSE-flexible
Evaluation results
| Method | Links | |
|---|---|---|
| DiSE-flexibleModel=LLaDA-1.5-8B2026.03 | 54.92 | |
| DiSE-flexibleModel=LLaDA-Instruct-8B2026.03 | 53.79 | |
| Baseline (Max Len)Model=LLaDA-1.5-8B2026.03 | 53.52 | |
| BaselineModel=LLaDA-1.5-8B2026.03 | 53.37 | |
| Baseline (Max Len)Model=LLaDA-Instruct-8B2026.03 | 52.38 | |
| BaselineModel=LLaDA-Instruct-8B2026.03 | 52.24 |