| Task Name | Dataset Name | SOTA Result | Trend | |
|---|---|---|---|---|
| Watermark Extraction | Legal domain dataset | WER37.5 | 48 | |
| Legal Knowledge Evaluation | Legal Domain CaseHOLD, MMLU-L | CaseHOLD Score79 | 26 | |
| Summarization | Legal domain (20 random samples) | Coherence2.7 | 4 | |
| Legal Reasoning | Legal domain (val) | Correctness3.1 | 2 |