Legal interpretation drafting on Expert-verified regulatory chunks (blind evaluation set)
30WinsHUMBR
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| HUMBRComparison=vs. Human Expert*2026.04 | 30 | 7 | 81 | |
| HUMBRComparison=vs. Universal Self-Consistency*2026.04 | 17 | 2 | 89.5 | |
| Universal Self-ConsistencyComparison=vs. Human2026.04 | 11 | 6 | 64.7 |