Common-sense Reasoning on ARC Challenge (Pass@1, FLOPS)
86.73Pass@1MFS (Ours)
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| MFS (Ours)Backbone=LLaMA3.1-8B-Instruct2026.01 | 86.73 | — | |
| ϕ-DecodingBackbone=LLaMA3.1-8B-Instruct2026.01 | 85.41 | — | |
| Predictive DecodingBackbone=LLaMA3.1-8B-Instruct2026.01 | 84.56 | — | |
| Guided DecodingBackbone=LLaMA3.1-8B-Instruct2026.01 | 81.74 | — | |
| Tree-of-ThoughtsBackbone=LLaMA3.1-8B-Instruct2026.01 | 80.72 | — | |
| MCTSBackbone=LLaMA3.1-8B-Instruct2026.01 | 79.95 | — | |
| MFS (Ours)Backbone=Mistral-v0.3-7B-Instruct2026.01 | 79.95 | — | |
| ϕ-DecodingBackbone=Mistral-v0.3-7B-Instruct2026.01 | 78.16 | — | |
| MCTSBackbone=Mistral-v0.3-7B-Instruct2026.01 | 74.74 | — | |
| Tree-of-ThoughtsBackbone=Mistral-v0.3-7B-Instruct2026.01 | 73.63 | — | |
| Guided DecodingBackbone=Mistral-v0.3-7B-Instruct2026.01 | 73.55 | — | |
| Predictive DecodingBackbone=Mistral-v0.3-7B-Instruct2026.01 | 73.55 | — | |
| Auto-Regressive (CoT)Backbone=Mistral-v0.3-7B-Instruct2026.01 | 69.54 | — | |
| Auto-Regressive (CoT)Backbone=LLaMA3.1-8B-Instruct2026.01 | 58.91 | — |