General Reasoning Evaluation on Average (AVG)
71.25AccuracyVanilla
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| VanillaBackbone=Llama3.1-8B2026.04 | 71.25 | 10.42 | 3,571 | 10.5 | |
| LThinker_thoBackbone=Llama3.1-8B2026.04 | 66.85 | 10.34 | 1,055 | 2.7 | |
| Vanilla + H2OBackbone=Llama3.1-8B2026.04 | 66.79 | 17.83 | 984 | 3.2 | |
| LThinker_tokBackbone=Llama3.1-8B2026.04 | 63.13 | 11.02 | 1,826 | 4.8 | |
| VanillaBackbone=Qwen2.5-7B2026.04 | 62.9 | 13.68 | 4,336 | 16.6 | |
| LThinker_thoBackbone=Qwen2.5-7B2026.04 | 62.8 | 10.17 | 1,289 | 3.7 | |
| CoTBackbone=Qwen2.5-7B2026.04 | 61.21 | 1.18 | 675 | 0.2 | |
| Vanilla + H2OBackbone=Qwen2.5-7B2026.04 | 61.12 | 20.61 | 972 | 4.4 | |
| CoTBackbone=Llama3.1-8B2026.04 | 60.54 | 1.61 | 790 | 0.4 | |
| AnLLMBackbone=Llama3.1-8B2026.04 | 59.04 | 12.86 | 659 | 1.9 | |
| LThinker_tokBackbone=Qwen2.5-7B2026.04 | 58.89 | 10.62 | 1,838 | 7.2 | |
| Distill-R1Backbone=Qwen2.5-7B2026.04 | 54.03 | 8.36 | 3,003 | 11.3 | |
| Distill-R1Backbone=Llama3.1-8B2026.04 | 52.12 | 2.91 | 1,332 | 4.4 | |
| AnLLMBackbone=Qwen2.5-7B2026.04 | 51.92 | 12.14 | 1,592 | 4.6 | |
| Vanilla + SepLLMBackbone=Llama3.1-8B2026.04 | 17.11 | 34.98 | 1,024 | 8 | |
| Vanilla + SepLLMBackbone=Qwen2.5-7B2026.04 | 12.32 | 36.32 | 1,024 | 8.9 |