First-Order Logic Reasoning on LogicNLI
76.6Pass@1SFT w/ Long CoT
Evaluation results
| Method | Links | ||
|---|---|---|---|
| SFT w/ Long CoTMethod Category=BEHAVIORAL CLONING, Reasoning Strategy=Long CoT2025.06 | 76.6 | — | |
| Dr. GRPOMethod Category=ADVANCED RLVRS2025.06 | 75.4 | — | |
| SFT w/o CoTMethod Category=BEHAVIORAL CLONING, Reasoning Strategy=w/o CoT2025.06 | 74.8 | — | |
| SFT w/ Short CoTMethod Category=BEHAVIORAL CLONING, Reasoning Strategy=Short CoT2025.06 | 73.8 | — | |
| RULEREASONER-8BMethod Category=RULEREASONER (Ours), Model Scale=8B2025.06 | 70.4 | 0.1 | |
| GRPOMethod Category=ADVANCED RLVRS2025.06 | 66.6 | — | |
| Easy-to-hard RLMethod Category=CURRICULUM LEARNING2025.06 | 66.6 | — | |
| RULEREASONER-4BMethod Category=RULEREASONER (Ours), Model Scale=4B2025.06 | 66.6 | 0.4 | |
| DAPOMethod Category=ADVANCED RLVRS2025.06 | 65.8 | — | |
| Data-balance RLMethod Category=CURRICULUM LEARNING2025.06 | 64.4 | — | |
| ADARFTMethod Category=CURRICULUM LEARNING2025.06 | 64.4 | — | |
| OpenAI o1Method Category=FRONTIER REASONERS2025.06 | 60 | — | |
| Claude-3.7-SonnetMethod Category=FRONTIER REASONERS2025.06 | 58 | — | |
| RGFBMethod Category=PRIOR RBRS2025.06 | 55 | — | |
| OpenAI o3-miniMethod Category=FRONTIER REASONERS2025.06 | 55 | — | |
| HtTMethod Category=PRIOR RBRS2025.06 | 54 | — | |
| Chain-of-LogicMethod Category=PRIOR RBRS2025.06 | 54 | — | |
| DeepSeek-R1Method Category=FRONTIER REASONERS2025.06 | 49 | — |