Mathematical Reasoning on Aggregate (AMC23, AIME24, MATH-500, Minerva, Olympiad)
68.2Intelligence Per Token (IPT)TRiMS
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| TRiMSBase Model=DeepSeek-R1-Distill-Qwen, Model Scale=7B2026.03 | 68.2 | 2.5 | -84.2 | |
| ExpThinkBackbone=Qwen3-8B2026.05 | 51.04 | 5.33 | -77.18 | |
| TRiMSBase Model=DeepSeek-R1-Distill-Qwen, Model Scale=1.5B2026.03 | 45.1 | 6.8 | -83.5 | |
| ExpThinkBackbone=DeepSeek-R1-Distill-Qwen-7B2026.05 | 32.84 | 1.8 | -62.11 | |
| LC-R1Base Model=DeepSeek-R1-Distill-Qwen, Model Scale=7B2026.03 | 28.4 | -0.3 | -58.1 | |
| LaserBase Model=DeepSeek-R1-Distill-Qwen, Model Scale=7B2026.03 | 27.8 | 8.8 | -57.6 | |
| LC-R1Backbone=DeepSeek-R1-Distill-Qwen-7B2026.05 | 23.47 | -5.94 | -48.88 | |
| LaserBackbone=DeepSeek-R1-Distill-Qwen-7B2026.05 | 23.3 | 2.62 | -50.17 | |
| ExpThinkBackbone=DeepSeek-R1-Distill-Qwen-1.5B2026.05 | 23.29 | 4.53 | -65.51 | |
| JETBase Model=DeepSeek-R1-Distill-Qwen, Model Scale=7B2026.03 | 21.1 | 2.1 | -43 | |
| AdaptThinkBackbone=DeepSeek-R1-Distill-Qwen-7B2026.05 | 19.52 | -0.88 | -33.78 | |
| AdaptThinkBase Model=DeepSeek-R1-Distill-Qwen, Model Scale=7B2026.03 | 19.3 | 4.7 | -35.2 | |
| AutoThinkBackbone=DeepSeek-R1-Distill-Qwen-7B2026.05 | 18.63 | 1.21 | -35.09 | |
| AdaptThinkBase Model=DeepSeek-R1-Distill-Qwen, Model Scale=1.5B2026.03 | 18.5 | 2.2 | -57.4 | |
| AdaptThinkBackbone=DeepSeek-R1-Distill-Qwen-1.5B2026.05 | 18.04 | 0.73 | -57.28 | |
| AutoThinkBase Model=DeepSeek-R1-Distill-Qwen, Model Scale=7B2026.03 | 17.9 | 2.6 | -36.4 | |
| JETBase Model=DeepSeek-R1-Distill-Qwen, Model Scale=1.5B2026.03 | 16.7 | -3.5 | -60.1 | |
| LaserBase Model=DeepSeek-R1-Distill-Qwen, Model Scale=1.5B2026.03 | 15.8 | 8.4 | -48.2 | |
| AutoThinkBase Model=DeepSeek-R1-Distill-Qwen, Model Scale=1.5B2026.03 | 14.9 | 1.9 | -45.3 | |
| LC-R1Backbone=DeepSeek-R1-Distill-Qwen-1.5B2026.05 | 14.49 | -4.19 | -50.01 | |
| JETBackbone=DeepSeek-R1-Distill-Qwen-7B2026.05 | 14.35 | 1.87 | -19.19 | |
| LC-R1Base Model=DeepSeek-R1-Distill-Qwen, Model Scale=1.5B2026.03 | 13.6 | -11.9 | -49.9 | |
| LaserBackbone=DeepSeek-R1-Distill-Qwen-1.5B2026.05 | 13.35 | 4.78 | -42.39 | |
| AutoThinkBackbone=DeepSeek-R1-Distill-Qwen-1.5B2026.05 | 13.25 | -1.3 | -43.5 | |
| VanillaBase Model=DeepSeek-R1-Distill-Qwen, Model Scale=7B2026.03 | 11.2 | — | — | |
| VanillaBackbone=DeepSeek-R1-Distill-Qwen-7B2026.05 | 11.15 | — | — | |
| JETBackbone=DeepSeek-R1-Distill-Qwen-1.5B2026.05 | 10.86 | 4.37 | -29.25 | |
| VanillaBackbone=Qwen3-8B2026.05 | 8.64 | — | — | |
| VanillaBackbone=DeepSeek-R1-Distill-Qwen-1.5B2026.05 | 7.23 | — | — | |
| VanillaBase Model=DeepSeek-R1-Distill-Qwen, Model Scale=1.5B2026.03 | 7.2 | — | — |