Out-of-Domain Generalization on LiveCodeBench, GPQA-Diamond, and MMLU Average
55.39IPTExpThink
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| ExpThinkBackbone=DeepSeek-R1-Distill-Qwen-7B2026.05 | 55.39 | 14.74 | -50.86 | |
| LC-R1Backbone=DeepSeek-R1-Distill-Qwen-7B2026.05 | 36.75 | -0.25 | -35.62 | |
| LaserBackbone=DeepSeek-R1-Distill-Qwen-7B2026.05 | 33.35 | 12.34 | -50.17 | |
| ExpThinkBackbone=DeepSeek-R1-Distill-Qwen-1.5B2026.05 | 27.26 | 26.44 | -50.1 | |
| AdaptThinkBackbone=DeepSeek-R1-Distill-Qwen-7B2026.05 | 26.91 | 4.54 | -19.78 | |
| AutoThinkBackbone=DeepSeek-R1-Distill-Qwen-7B2026.05 | 25.79 | 4.11 | -20.37 | |
| JETBackbone=DeepSeek-R1-Distill-Qwen-7B2026.05 | 23.95 | 7 | -9.29 | |
| VanillaBackbone=DeepSeek-R1-Distill-Qwen-7B2026.05 | 20.78 | — | — | |
| AdaptThinkBackbone=DeepSeek-R1-Distill-Qwen-1.5B2026.05 | 17.71 | 24.52 | -30.48 | |
| LC-R1Backbone=DeepSeek-R1-Distill-Qwen-1.5B2026.05 | 17.18 | 7.83 | -43.67 | |
| LaserBackbone=DeepSeek-R1-Distill-Qwen-1.5B2026.05 | 17.16 | 18.21 | -26.92 | |
| JETBackbone=DeepSeek-R1-Distill-Qwen-1.5B2026.05 | 16.93 | 20.02 | -17.35 | |
| AutoThinkBackbone=DeepSeek-R1-Distill-Qwen-1.5B2026.05 | 14.92 | 19.72 | -26.5 | |
| VanillaBackbone=DeepSeek-R1-Distill-Qwen-1.5B2026.05 | 9.01 | — | — |