General Knowledge on MMLU (pass@1)
72.48pass@1SPINE
Evaluation Results
| Method | Links | |
|---|---|---|
| SPINEBase Model=Qwen3-1.7B2025.11 | 72.48 | |
| LMSIBase Model=Qwen3-1.7B2025.11 | 71.74 | |
| TTRLBase Model=Qwen3-1.7B2025.11 | 71.19 | |
| SEALONGBase Model=Qwen3-1.7B2025.11 | 69.36 | |
| w/o MergingModels=Code2026.02 | 68.39 | |
| Self-ConsistencyBase Model=Qwen3-1.7B2025.11 | 65.39 | |
| Task ArithmeticModels=Fuse2026.02 | 65.17 | |
| Dare Task ArithmeticModels=Fuse2026.02 | 64.85 | |
| Dare Ties MergingModels=Fuse2026.02 | 63.76 | |
| SCF-RKLModels=Fuse2026.02 | 63.47 | |
| VSBToken budget=3, Number of shots=0, Backbone=LLaDA-8B2026.04 | 63.11 | |
| AdaBlockToken budget=3, Number of shots=02026.04 | 62.98 | |
| VSBToken budget=3, Number of shots=0, Backbone=LLaDA-1.52026.04 | 62.94 | |
| Blockwise-SFTToken budget=3, Number of shots=02026.04 | 62.81 | |
| D2FToken budget=3, Number of shots=02026.04 | 62.71 | |
| DAEDALToken budget=3, Number of shots=02026.04 | 62.58 | |
| LLaDA-1.5Token budget=3, Number of shots=02026.04 | 62.52 | |
| LLaDA-8BToken budget=3, Number of shots=02026.04 | 62.49 | |
| w/o MergingModels=Math2026.02 | 61.83 | |
| dLLM-VarToken budget=3, Number of shots=02026.04 | 61.71 | |
| Ties MergingModels=Fuse2026.02 | 58.94 | |
| No adaptationBase Model=Qwen3-1.7B2025.11 | 58.16 | |
| SCEModels=Fuse2026.02 | 57.12 | |
| Mistral-7B-Instruct-v0.2Models=base02026.02 | 43.23 | |
| Dare Ties MergingModels=Fuse2026.02 | 41.78 | |
| SCF-RKLModels=Fuse2026.02 | 38.79 | |
| Task ArithmeticModels=Fuse2026.02 | 26.29 | |
| Dare Task ArithmeticModels=Fuse2026.02 | 26.29 | |
| Ties MergingModels=Fuse2026.02 | 19.37 | |
| SCEModels=Fuse2026.02 | 19.28 | |
| MathCoder2-Mistral-7BModels=base12026.02 | 19.14 |