Legal Evaluation on LexEval
71.1AccuracyLawThinker
Evaluation Results
| Method | Links | |
|---|---|---|
| LawThinkerBackbone=Qwen3-32B2026.02 | 71.1 | |
| Search-o1Backbone=Qwen3-32B2026.02 | 69 | |
| ReActBackbone=Qwen3-32B2026.02 | 68.6 | |
| Plan-and-ExecuteBackbone=Qwen3-32B2026.02 | 68.1 | |
| Direct Reasoning2026.02 | 67.1 | |
| Plan-and-SolveBackbone=Qwen3-32B2026.02 | 66.6 | |
| SYNDATAModel=Qwen3-4B, Data Size=40002026.05 | 48 | |
| SYNDATAModel=Qwen2.5-7B, Data Size=40002026.05 | 46.15 | |
| SYNDATAModel=Qwen3-4B, Data Size=20002026.05 | 45.44 | |
| SYNDATAModel=Qwen2.5-7B, Data Size=20002026.05 | 42.8 | |
| SYNDATAModel=Qwen3-4B, Data Size=10002026.05 | 37.9 | |
| SYNDATAModel=Qwen2.5-7B, Data Size=10002026.05 | 33.87 | |
| Base ModelModel=Qwen3-4B, Data Size=-2026.05 | 31.48 | |
| Base ModelModel=Qwen2.5-7B, Data Size=-2026.05 | 19.8 |