Evidence Re-Ranking on Evidence Re-Ranking
66.13Rel-F1STL Classification (CLS)
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| STL Classification (CLS)Learning Paradigm=Single-Task Learning, Backbone=Qwen3-4B, Configuration=Classification Head2026.01 | 66.13 | 96.66 | 81.39 | 94.24 | |
| MTL CLSLearning Paradigm=Multi-Task Learning, Backbone=Qwen3-4B, Configuration=Classification Head2026.01 | 64.15 | 96.63 | 80.39 | 94.05 | |
| XLM-RoBERTa-LargeBackbone=XLM-RoBERTa-Large2026.01 | 43.9 | 92.5 | 68.2 | 84.9 | |
| STL Instruction TuningLearning Paradigm=Single-Task Learning, Backbone=Qwen3-4B, Configuration=Instruction Tuning2026.01 | 42.4 | 95.8 | 46.1 | 91.6 | |
| MTL Instruction TuningLearning Paradigm=Multi-Task Learning, Backbone=Qwen3-4B, Configuration=Instruction Tuning2026.01 | 42.2 | 94.8 | 68.5 | 90.6 | |
| Zero-shotMode=Zero-shot2026.01 | 39.7 | 92.1 | 65.9 | 87.9 | |
| Few-shot ICLMode=Few-shot In-Context Learning2026.01 | 39.7 | 93.2 | 66.5 | 89 | |
| MTL CLMLearning Paradigm=Multi-Task Learning, Backbone=Qwen3-4B, Configuration=Causal Language Model2026.01 | 24.82 | 93.87 | 59.35 | 88.39 | |
| STL Causal LMLearning Paradigm=Single-Task Learning, Backbone=Qwen3-4B, Configuration=Causal Language Model2026.01 | 12.82 | 74.41 | 43.61 | 69.52 |