Rhetorical Role Labeling on PUBMED
100Macro F1Gold Prototypes
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Gold PrototypesEvaluation protocol=optimal prototype post-hoc for each sentence2026.03 | 100 | 100 | |
| PCMmethod_type=Prototype-Based2026.03 | 88.86 | 92.91 | |
| PBR2026.03 | 88.86 | 92.91 | |
| Mind2026.03 | 87.67 | 91.86 | |
| PBRmethod_type=Prototype-Based2026.03 | 87.42 | 92.06 | |
| PCMSampling strategy=Supervised Sampling2026.03 | 87.42 | 92.06 | |
| PCMSampling strategy=Random Sampling2026.03 | 87.24 | 91.94 | |
| PCMSampling strategy=Full Sampling2026.03 | 87.19 | 91.89 | |
| Baseline (HSLN)2026.03 | 87.01 | 91.09 | |
| Meta-Llama3-8Badaptation_method=QLoRA2026.03 | 82.51 | 87.63 | |
| Mistral-7Badaptation_method=QLoRA2026.03 | 81.86 | 87.2 | |
| Qwen3-8Badaptation_method=QLoRA2026.03 | 81.73 | 87.26 | |
| DeepSeek-70Badaptation_method=QLoRA2026.03 | 81.03 | 86.67 |