Legal Reasoning on JEC-QA Legal (val)
54.4Accuracy+CLARITY
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| CLARITYModel Size=7B2025.10 | 54.4 | 94.1 | 56.4 | |
| Refine-then-Monitor2025.10 | 52.3 | 93.8 | 56.3 | |
| Data Reformulation2025.10 | 47.4 | 75.3 | 56.9 | |
| Standard RL2025.10 | 45 | 69.3 | 58.5 | |
| PRM RL2025.10 | 42 | 68.2 | 54.7 | |
| GPT-4o2025.10 | 39.5 | 96.6 | 39.9 | |
| CLARITYModel Size=3B2025.10 | 33.9 | 97.1 | 34.2 | |
| Qwen2.5Model Size=7B2025.10 | 28.5 | 92.3 | 29.3 | |
| Qwen2.5Model Size=3B2025.10 | 17.5 | 74.3 | 19.3 |