Mathematical Reasoning on LiveMathBench 202505
32.17AccuracyOriginal
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| OriginalPrompt=step, Backbone=Qwen3-8B2026.02 | 32.17 | 17,493 | 18 | |
| D-SFT->D-RL3k->D-RL6kPrompt=adaptive, Backbone=Qwen3-8B2026.02 | 31.83 | 10,341 | 31 | |
| D-SFT->D-RL6kPrompt=step, Backbone=Qwen3-8B2026.02 | 31.33 | 13,957 | 22 | |
| D-SFT->D-RL3k->D-RL6kPrompt=step, Backbone=Qwen3-8B2026.02 | 30.83 | 13,980 | 22 | |
| D-SFT->D-RL6kPrompt=adaptive, Backbone=Qwen3-8B2026.02 | 30.67 | 10,741 | 28 | |
| ThinkPrune(6k)Prompt=step, Backbone=Qwen3-8B2026.02 | 28.67 | 11,043 | 26 | |
| ThinkPrune(6k->3k)Prompt=step, Backbone=Qwen3-8B2026.02 | 28.67 | 9,333 | 31 | |
| D-SFT->D-RL6kPrompt=draft, Backbone=Qwen3-8B2026.02 | 26.33 | 6,949 | 37 | |
| D-SFT->D-RL3kPrompt=draft, Backbone=Qwen3-8B2026.02 | 22.33 | 5,002 | 45 | |
| D-SFT->D-RL3k->D-RL6kPrompt=draft, Backbone=Qwen3-8B2026.02 | 22.33 | 6,476 | 35 | |
| D-SFTPrompt=draft, Backbone=Qwen3-8B2026.02 | 12.5 | 3,732 | 33 |