Document Grounding on Hebrew Verdict Decisions human-annotated gold labels
78.1Macro F1 (Partial)Doc-Structure-Parser
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Doc-Structure-Parser2024.02 | 78.1 | 63.7 | 82.8 | 64.2 | |
| Randomaveraging=100 different random seeds2024.02 | 21.6 | 6.4 | 16.6 | 3.2 | |
| Most frequent classclass_strategy=prosecution evidence2024.02 | 8.3 | 0 | 18.7 | 0 |