Mathematical Reasoning on GSM8K n=200 (pilot)
94AccuracyMrsd (3-round)
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Mrsd (3-round)Model=Qwen3-8B, Decoding Strategy=greedy decoding2026.05 | 94 | 235 | |
| Nothink@B1Model=Qwen3-8B, Decoding Strategy=greedy decoding2026.05 | 89 | 140 | |
| Town@1024Model=Qwen3-8B, Decoding Strategy=greedy decoding2026.05 | 89 | 180 | |
| Think@BthinkModel=Qwen3-8B, Decoding Strategy=greedy decoding2026.05 | 56.9 | 460 |