Legal Reasoning (Top-1 Accuracy) on LawBench
70.1Top-1 AccuracySIA-W+H
Evaluation Results
| Method | Links | |
|---|---|---|
| SIA-W+Hharness + weights=true2026.05 | 70.1 | |
| SIA-Hharness only=true2026.05 | 50 | |
| Previous SOTA2026.05 | 45 | |
| gpt-oss-120bstatus=initial scaffold2026.05 | 13.5 |
Loading the SOTA2 catalog…