Mathematical Reasoning on AMC 2023 (Custom Metrics)
100AccuracyDEER
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| DEERModel Base Pair=Qwen3-4B-2507-Thinking/Qwen3-4B-2507-Instruct, Sampling Rounds=8x2026.01 | 100 | 10,084 | 91 | 100 | |
| TAModel Base Pair=Qwen3-4B-2507-Thinking/Qwen3-4B-2507-Instruct, Sampling Rounds=8x2026.01 | 100 | 7,760 | 70 | 100 | |
| MI-07Model Base Pair=Qwen3-4B-2507-Thinking/Qwen3-4B-2507-Instruct, Sampling Rounds=8x2026.01 | 100 | 7,501 | 68 | 100 | |
| RoutingModel Base Pair=Qwen3-4B-2507-Thinking/Qwen3-4B-2507-Instruct, Sampling Rounds=8x2026.01 | 100 | 10,356 | 93 | 85 | |
| ThinkingModel Base Pair=Qwen3-4B-2507-Thinking/Qwen3-4B-2507-Instruct, Sampling Rounds=8x2026.01 | 99.4 | 11,076 | 100 | 100 | |
| DAMI-PrefModel Base Pair=Qwen3-4B-2507-Thinking/Qwen3-4B-2507-Instruct, Sampling Rounds=8x2026.01 | 98.8 | 7,872 | 71 | 86 | |
| DAMI-ConfModel Base Pair=Qwen3-4B-2507-Thinking/Qwen3-4B-2507-Instruct, Sampling Rounds=8x2026.01 | 98.8 | 5,891 | 53 | 55 | |
| MI-03Model Base Pair=Qwen3-4B-2507-Thinking/Qwen3-4B-2507-Instruct, Sampling Rounds=8x2026.01 | 96.2 | 3,381 | 31 | 0 | |
| MI-05Model Base Pair=Qwen3-4B-2507-Thinking/Qwen3-4B-2507-Instruct, Sampling Rounds=8x2026.01 | 96.2 | 5,138 | 46 | 61 | |
| TIESModel Base Pair=Qwen3-4B-2507-Thinking/Qwen3-4B-2507-Instruct, Sampling Rounds=8x2026.01 | 95.6 | 3,012 | 27 | 0 | |
| DAMI-PromptModel Base Pair=Qwen3-4B-2507-Thinking/Qwen3-4B-2507-Instruct, Sampling Rounds=8x2026.01 | 95.6 | 4,715 | 43 | 30 | |
| DAMI-PredModel Base Pair=Qwen3-4B-2507-Thinking/Qwen3-4B-2507-Instruct, Sampling Rounds=8x2026.01 | 95 | 3,477 | 31 | 0 | |
| InstructModel Base Pair=Qwen3-4B-2507-Thinking/Qwen3-4B-2507-Instruct, Sampling Rounds=8x2026.01 | 93.8 | 3,034 | 27.4 | 0 | |
| DAMI-PrefModel Base Pair=DeepSeek-R1-Distill-Qwen-7B/Qwen2.5-Math-7B, Sampling Rounds=8x2026.01 | 88.8 | 5,819 | 68 | 88 | |
| ThinkingModel Base Pair=DeepSeek-R1-Distill-Qwen-7B/Qwen2.5-Math-7B, Sampling Rounds=8x2026.01 | 87.5 | 8,600 | 100 | 100 | |
| DAMI-ConfModel Base Pair=DeepSeek-R1-Distill-Qwen-7B/Qwen2.5-Math-7B, Sampling Rounds=8x2026.01 | 87.5 | 5,299 | 62 | 78 | |
| MI-07Model Base Pair=DeepSeek-R1-Distill-Qwen-7B/Qwen2.5-Math-7B, Sampling Rounds=8x2026.01 | 85 | 4,867 | 57 | 92 | |
| DEERModel Base Pair=DeepSeek-R1-Distill-Qwen-7B/Qwen2.5-Math-7B, Sampling Rounds=8x2026.01 | 82.5 | 7,398 | 86 | 100 | |
| TAModel Base Pair=DeepSeek-R1-Distill-Qwen-7B/Qwen2.5-Math-7B, Sampling Rounds=8x2026.01 | 78.1 | 4,319 | 50 | 96 | |
| MI-05Model Base Pair=DeepSeek-R1-Distill-Qwen-7B/Qwen2.5-Math-7B, Sampling Rounds=8x2026.01 | 72.5 | 4,462 | 52 | 8 | |
| DAMI-PredModel Base Pair=DeepSeek-R1-Distill-Qwen-7B/Qwen2.5-Math-7B, Sampling Rounds=8x2026.01 | 62.5 | 3,103 | 36 | 20 | |
| TIESModel Base Pair=DeepSeek-R1-Distill-Qwen-7B/Qwen2.5-Math-7B, Sampling Rounds=8x2026.01 | 58.1 | 1,263 | 15 | 0 | |
| MI-03Model Base Pair=DeepSeek-R1-Distill-Qwen-7B/Qwen2.5-Math-7B, Sampling Rounds=8x2026.01 | 57.5 | 1,817 | 21 | 0 | |
| DAMI-PromptModel Base Pair=DeepSeek-R1-Distill-Qwen-7B/Qwen2.5-Math-7B, Sampling Rounds=8x2026.01 | 55 | 1,697 | 20 | 0 | |
| RoutingModel Base Pair=DeepSeek-R1-Distill-Qwen-7B/Qwen2.5-Math-7B, Sampling Rounds=8x2026.01 | 50 | 1,425 | 17 | 5 | |
| InstructModel Base Pair=DeepSeek-R1-Distill-Qwen-7B/Qwen2.5-Math-7B, Sampling Rounds=8x2026.01 | 42.5 | 1,463 | — | 0 |