Rhetorical Role Labeling on CS-ABSTRACTS
99.84Weighted F1Gold Prototypes
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Gold PrototypesEvaluation protocol=optimal prototype post-hoc for each sentence2026.03 | 99.84 | — | 99.66 | |
| PCMmethod_type=Prototype-Based2026.03 | 78.09 | 71.1 | — | |
| PBR2026.03 | 78.09 | — | 71.1 | |
| Mind2026.03 | 76.91 | — | 69.19 | |
| PCMSampling strategy=Full Sampling2026.03 | 76.66 | — | 69.84 | |
| PCMSampling strategy=Random Sampling2026.03 | 76.3 | — | 69.12 | |
| PBRmethod_type=Prototype-Based2026.03 | 75.46 | 68.69 | — | |
| PCMSampling strategy=Supervised Sampling2026.03 | 75.46 | — | 68.69 | |
| Baseline (HSLN)2026.03 | 75.08 | — | 68.55 | |
| Qwen3-8Badaptation_method=QLoRA2026.03 | 74.93 | 65.93 | — | |
| Meta-Llama3-8Badaptation_method=QLoRA2026.03 | 74.59 | 67.39 | — | |
| DeepSeek-70Badaptation_method=QLoRA2026.03 | 72.6 | 64.92 | — | |
| Mistral-7Badaptation_method=QLoRA2026.03 | 70.61 | 62.24 | — |