Legal Reasoning on LegalBench Learned Hands Courts
75.5AccuracyMAD
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| MADBackbone=Gemini-2.5-Flash2026.02 | 75.5 | 76.8 | 75.5 | 75.2 | |
| ACALBackbone=Gemini-2.5-Flash2026.02 | 75.5 | 76.4 | 75.5 | 75.3 | |
| RAGBackbone=Gemini-2.5-Flash2026.02 | 75 | 78.1 | 75 | 74.3 | |
| CoTBackbone=Gemini-2.5-Flash2026.02 | 72.3 | 74.1 | 72.6 | 71.3 | |
| ACALBackbone=Gemini-2.5-Flash-Lite2026.02 | 70.8 | 71.2 | 70.8 | 70.7 | |
| RAGBackbone=Gemini-2.5-Flash-Lite2026.02 | 70.3 | 70.5 | 70.3 | 70.3 | |
| MADBackbone=Gemini-2.5-Flash-Lite2026.02 | 69.8 | 69.8 | 69.8 | 69.8 | |
| CoTBackbone=Gemini-2.5-Flash-Lite2026.02 | 69.3 | 69.3 | 69.2 | 69.2 | |
| SPBackbone=Gemini-2.5-Flash2026.02 | 65.1 | 72.5 | 65.1 | 62 | |
| SPBackbone=Gemini-2.5-Flash-Lite2026.02 | 57.8 | 63.1 | 57.8 | 53.1 |