First-Order Logic Reasoning on FOLIO
84.7Pass@1 Success RateRULEREASONER-8B
Evaluation results
| Method | Links | ||
|---|---|---|---|
| RULEREASONER-8BMethod Category=RULEREASONER (Ours), Model Scale=8B2025.06 | 84.7 | 0.6 | |
| SFT w/ Long CoTMethod Category=BEHAVIORAL CLONING, Reasoning Strategy=Long CoT2025.06 | 83.4 | — | |
| SFT w/ Short CoTMethod Category=BEHAVIORAL CLONING, Reasoning Strategy=Short CoT2025.06 | 82.9 | — | |
| ADARFTMethod Category=CURRICULUM LEARNING2025.06 | 81.8 | — | |
| DAPOMethod Category=ADVANCED RLVRS2025.06 | 80.9 | — | |
| Chain-of-LogicMethod Category=PRIOR RBRS2025.06 | 80 | — | |
| Easy-to-hard RLMethod Category=CURRICULUM LEARNING2025.06 | 78.9 | — | |
| RULEREASONER-4BMethod Category=RULEREASONER (Ours), Model Scale=4B2025.06 | 78.9 | 0.8 | |
| OpenAI o1Method Category=FRONTIER REASONERS2025.06 | 77 | — | |
| Data-balance RLMethod Category=CURRICULUM LEARNING2025.06 | 76.8 | — | |
| Claude-3.7-SonnetMethod Category=FRONTIER REASONERS2025.06 | 74.7 | — | |
| RGFBMethod Category=PRIOR RBRS2025.06 | 74 | — | |
| OpenAI o3-miniMethod Category=FRONTIER REASONERS2025.06 | 74 | — | |
| Dr. GRPOMethod Category=ADVANCED RLVRS2025.06 | 73.9 | — | |
| SFT w/o CoTMethod Category=BEHAVIORAL CLONING, Reasoning Strategy=w/o CoT2025.06 | 73.4 | — | |
| DeepSeek-R1Method Category=FRONTIER REASONERS2025.06 | 72.7 | — | |
| GRPOMethod Category=ADVANCED RLVRS2025.06 | 72.3 | — | |
| HtTMethod Category=PRIOR RBRS2025.06 | 71 | — |