Mathematical Reasoning on MATH (test) (Accuracy + # Tokens + Time)
53.6AccuracyDebateOCR
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| DebateOCRBackbone Model=Pixtral-12B, Debate Rounds=5, Agents=32026.01 | 53.6 | 4.7 | 113.9 | |
| TS-MADBackbone Model=Pixtral-12B, Debate Rounds=5, Agents=32026.01 | 51.8 | 20.9 | 271.4 | |
| T-MADBackbone Model=Pixtral-12B, Debate Rounds=5, Agents=32026.01 | 49.2 | 70.1 | 256.4 | |
| DebateOCRBackbone Model=Llama-3.2-11B, Debate Rounds=5, Agents=32026.01 | 47.1 | 4.7 | 106.8 | |
| DebateOCRBackbone Model=Qwen2.5-VL-7B, Debate Rounds=5, Agents=32026.01 | 46.8 | 4.7 | 64 | |
| TS-MADBackbone Model=Llama-3.2-11B, Debate Rounds=5, Agents=32026.01 | 46.7 | 22.3 | 254.4 | |
| T-MADBackbone Model=Llama-3.2-11B, Debate Rounds=5, Agents=32026.01 | 45.2 | 74.8 | 240.5 | |
| TS-MADBackbone Model=Qwen2.5-VL-7B, Debate Rounds=5, Agents=32026.01 | 45.1 | 21.5 | 152.7 | |
| T-MADBackbone Model=Qwen2.5-VL-7B, Debate Rounds=5, Agents=32026.01 | 44.8 | 72 | 144.2 | |
| DebateOCRBackbone Model=InternVL-8B, Debate Rounds=5, Agents=32026.01 | 44.2 | 4.7 | 71.2 | |
| TS-MADBackbone Model=InternVL-8B, Debate Rounds=5, Agents=32026.01 | 43.1 | 19.8 | 169.6 | |
| T-MADBackbone Model=InternVL-8B, Debate Rounds=5, Agents=32026.01 | 42.5 | 66.3 | 160.3 |