Logical Reasoning on LogiQA (Pass@1, FLOPS)
48.61Pass@1 AccuracyMFS (Ours)
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| MFS (Ours)Backbone=LLaMA3.1-8B-Instruct2026.01 | 48.61 | — | |
| ϕ-DecodingBackbone=LLaMA3.1-8B-Instruct2026.01 | 48.39 | — | |
| Predictive DecodingBackbone=LLaMA3.1-8B-Instruct2026.01 | 46.7 | — | |
| Tree-of-ThoughtsBackbone=LLaMA3.1-8B-Instruct2026.01 | 45.93 | — | |
| Guided DecodingBackbone=LLaMA3.1-8B-Instruct2026.01 | 43.47 | — | |
| MFS (Ours)Backbone=Mistral-v0.3-7B-Instruct2026.01 | 43.16 | — | |
| ϕ-DecodingBackbone=Mistral-v0.3-7B-Instruct2026.01 | 43.01 | — | |
| MCTSBackbone=LLaMA3.1-8B-Instruct2026.01 | 42.7 | — | |
| Tree-of-ThoughtsBackbone=Mistral-v0.3-7B-Instruct2026.01 | 41.63 | — | |
| MCTSBackbone=Mistral-v0.3-7B-Instruct2026.01 | 40.71 | — | |
| Predictive DecodingBackbone=Mistral-v0.3-7B-Instruct2026.01 | 39.78 | — | |
| Auto-Regressive (CoT)Backbone=Mistral-v0.3-7B-Instruct2026.01 | 37.02 | — | |
| Guided DecodingBackbone=Mistral-v0.3-7B-Instruct2026.01 | 36.71 | — | |
| Auto-Regressive (CoT)Backbone=LLaMA3.1-8B-Instruct2026.01 | 33.33 | — |