Holistic formalism classification on MADON (test)
0.919Macro PrecisionMLP
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| MLPModel Category=Using gold-annotated arguments as features, Input Features=Gold-annotated arguments2025.12 | 0.919 | 0.915 | 0.917 | |
| MADON pipeline (Basic)Model Category=Multi-step pipelines, Pipeline Configuration=Basic2025.12 | 0.84 | 0.863 | 0.832 | |
| MADON pipeline (With Filtering)Model Category=Multi-step pipelines, Pipeline Configuration=With Filtering2025.12 | 0.838 | 0.872 | 0.828 | |
| ModernBERT-large-Czech-LegalModel Category=End-to-end models, Base Model=ModernBERT Large, Domain-specific Pre-training=Czech-Legal2025.12 | 0.824 | 0.82 | 0.822 | |
| Llama-3.1-8B-Czech-LegalModel Category=End-to-end models, Base Model=Llama 3.1 8B, Domain-specific Pre-training=Czech-Legal2025.12 | 0.799 | 0.775 | 0.78 | |
| Llama-3.1-8B-Czech-Legal + PEFTModel Category=End-to-end models, Base Model=Llama 3.1 8B, Domain-specific Pre-training=Czech-Legal, Adaptation=PEFT2025.12 | 0.796 | 0.629 | 0.603 | |
| Llama 3.1 8B BaseModel Category=End-to-end models, Base Model=Llama 3.1 8B2025.12 | 0.771 | 0.755 | 0.759 | |
| ModernBERT LargeModel Category=End-to-end models, Base Model=ModernBERT Large2025.12 | 0.741 | 0.739 | 0.738 | |
| Llama 3.1 8B Base + PEFTModel Category=End-to-end models, Base Model=Llama 3.1 8B, Adaptation=PEFT2025.12 | 0.725 | 0.605 | 0.576 | |
| Random baselineModel Category=Baselines2025.12 | 0.427 | 0.427 | 0.414 | |
| Majority baselineModel Category=Baselines2025.12 | 0.293 | 0.5 | 0.369 |