Mathematical Reasoning on AIME 24 (Acc, TR, s/q)
70AccuracyDEER
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| DEERBackbone=DeepSeek-R1-Distill-Qwen-14B2026.05 | 70 | 2.7 | 22.21 | |
| PUMABackbone=Llama-3.1-Nemotron-Nano-8B2026.05 | 70 | 12.3 | 8.71 | |
| PUMABackbone=DeepSeek-R1-Distill-Qwen-14B2026.05 | 66.7 | 16.3 | 15.62 | |
| Full CoTBackbone=Llama-3.1-Nemotron-Nano-8B2026.05 | 66.7 | 0 | 7.68 | |
| Full CoTBackbone=DeepSeek-R1-Distill-Qwen-14B2026.05 | 63.3 | 0 | 20.2 | |
| PUMABackbone=DeepSeek-R1-Distill-Qwen-7B2026.05 | 60 | 30.3 | 9.41 | |
| DynasorBackbone=DeepSeek-R1-Distill-Qwen-14B2026.05 | 56.7 | -3.5 | 12.54 | |
| DEERBackbone=Llama-3.1-Nemotron-Nano-8B2026.05 | 53.3 | -22.5 | 23.1 | |
| Full CoTBackbone=DeepSeek-R1-Distill-Qwen-7B2026.05 | 50 | 0 | 10.18 | |
| DynasorBackbone=Llama-3.1-Nemotron-Nano-8B2026.05 | 50 | -1 | 9.17 | |
| DEERBackbone=DeepSeek-R1-Distill-Qwen-7B2026.05 | 40 | 14.1 | 12.64 | |
| DynasorBackbone=DeepSeek-R1-Distill-Qwen-7B2026.05 | 23.3 | 21.7 | 11.86 |