Law
Benchmarks
| Task Name | Dataset Name | SOTA Result | Trend | |
|---|---|---|---|---|
| Law reasoning | Law | Accuracy70.25 | 27 | |
| Speculative Decoding | Law | Throughput (tokens/s)132.69 | 22 | |
| Counterfactual Explanations | Law | Validity100 | 18 | |
| Legal Reasoning | Law | LLM-as-judge Score34.4 | 13 | |
| Legal Reasoning | Law | Score26.52 | 13 | |
| Machine Translation | Law (test) | BLEU61.55 | 9 | |
| Multiple Choice Question Answering | Law | Accuracy45.4 | 8 | |
| Multi-class Classification | Law | Accuracy69.1 | 8 | |
| Machine Translation | Law (Ko-En) (test) | BLEU53.8 | 8 | |
| Machine Translation | Law De-En (test) | BLEU75.44 | 8 | |
| Machine Translation | Law All-domain datastore (test) | BLEU61.22 | 6 | |
| Legal Reasoning | Law (test) | Score45.29 | 5 | |
| Machine Translation | Law En-De out-of-domain WMT14 (test) | BLEU Score41.5 | 5 | |
| Within-distribution detection | Law Unsafe vs. Safe v3/v4 (holdout) | AUROC0.971 | 2 | |
| Machine Translation | Law multi-domain (test) | Decoding Speed (Tok/Sec)3,690.85 | 2 |
Showing 15 of 15 rows