Generation on LegalBench Rule-Application
59Exact MatchOracle (Upper Bound)
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Oracle (Upper Bound)2026.04 | 59 | 53.9 | |
| HUMBR2026.04 | 53.5 | 50.8 | |
| Self-Consistency (USC)2026.04 | 49.3 | 45.5 | |
| Greedy Decoding2026.04 | 44.4 | 42.9 |
Loading the SOTA2 catalog…