Mathematical Reasoning on MATH 500 (Acc, Delta_Up, Delta_Down)
89.8AccuracyAgentic-Verifier-Qwen3-4B
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Agentic-Verifier-Qwen3-4BRefinement Turn=32026.04 | 89.8 | 1.8 | 1.2 | |
| Agentic-Verifier-Qwen3-4BRefinement Turn=22026.04 | 89.2 | 2.4 | 1.2 | |
| DS-Distill-14BRefinement Turn=22026.04 | 87.4 | 5.8 | 1.4 | |
| DS-Distill-14BRefinement Turn=32026.04 | 85.8 | 2.2 | 3.8 | |
| Qwen3-4B-ThinkRefinement Turn=32026.04 | 84.8 | 2.4 | 2.2 | |
| Qwen3-4B-ThinkRefinement Turn=22026.04 | 84.6 | 6 | 1.4 | |
| Agentic-Verifier-Qwen3-4BRefinement Turn=12026.04 | 84.2 | 41.6 | 0.6 | |
| DS-Distill-14BRefinement Turn=12026.04 | 83 | 37.4 | 1.4 | |
| Qwen3-4B-ThinkRefinement Turn=12026.04 | 80 | 33.8 | 0.8 | |
| Mistral-Small-24B-InstructRefinement Turn=32026.04 | 67.8 | 4.4 | 3 | |
| Mistral-Small-24B-InstructRefinement Turn=22026.04 | 66.4 | 5.8 | 0.8 | |
| Qwen2.5-7B-InstructRefinement Turn=32026.04 | 66 | 4.2 | 3 | |
| Qwen2.5-7B-InstructRefinement Turn=22026.04 | 64.8 | 6.2 | 1.8 | |
| Mistral-Small-24B-InstructRefinement Turn=12026.04 | 61.4 | 15.6 | 1.2 | |
| Qwen2.5-7B-InstructRefinement Turn=12026.04 | 60.4 | 14 | 0.6 | |
| Llama3.1-8B-InstructRefinement Turn=32026.04 | 58.8 | 4 | 3.6 | |
| Llama3.1-8B-InstructRefinement Turn=22026.04 | 58.4 | 6 | 2.2 | |
| Llama3.1-8B-InstructRefinement Turn=12026.04 | 54.6 | 11 | 3.4 |