Mathematical Reasoning on AIME 2025 (Acc, #L)
39.33AccuracySPOT-stage2
Evaluation results
| Method | Links | ||
|---|---|---|---|
| SPOT-stage2Backbone=DeepSeek-R1-Distill-Qwen-7B2026.03 | 39.33 | 9,568 | |
| VanillaBackbone=DeepSeek-R1-Distill-Qwen-7B2026.03 | 36 | 11,357 | |
| DEERBackbone=DeepSeek-R1-Distill-Qwen-7B2026.03 | 35.67 | 10,124 | |
| CCoTBackbone=DeepSeek-R1-Distill-Qwen-7B2026.03 | 35.33 | 10,842 | |
| ConciseHintBackbone=DeepSeek-R1-Distill-Qwen-7B2026.03 | 34.67 | 9,822 | |
| Step EntropyBackbone=DeepSeek-R1-Distill-Qwen-7B2026.03 | 34.67 | 9,637 | |
| SPOT-stage1Backbone=DeepSeek-R1-Distill-Qwen-7B2026.03 | 34.67 | 9,471 | |
| LightThinkerBackbone=DeepSeek-R1-Distill-Qwen-7B2026.03 | 29.67 | 9,543 | |
| L1-MaxBackbone=DeepSeek-R1-Distill-Qwen-7B2026.03 | 24 | 4,636 | |
| Latent-SFTBackbone=DeepSeek-R1-Distill-Qwen-7B2026.03 | 10.67 | 2,074 | |
| COCONUTBackbone=DeepSeek-R1-Distill-Qwen-7B2026.03 | 6.67 | 1,135 | |
| CODIBackbone=DeepSeek-R1-Distill-Qwen-7B2026.03 | 6.33 | 1,836 |