Stance Detection
76.33Sup-F1STL Classification (CLS)
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| STL Classification (CLS)Learning Strategy=Single-Task Learning (STL), Model Strategy=Classification (CLS)2026.01 | 76.33 | 67.78 | 47.36 | 85.97 | 69.36 | 75.22 | |
| STL Instruction TuningLearning Strategy=Single-Task Learning (STL), Model Strategy=Instruction Tuning2026.01 | 74.9 | 69.5 | 38.3 | 86.5 | 67.3 | 75.8 | |
| MTL Instruction TuningLearning Strategy=Multi-Task Learning (MTL), Model Strategy=Instruction Tuning2026.01 | 74.8 | 68.2 | 45.4 | 85.6 | 68.5 | 74.6 | |
| XLM-RoBERTa-LargeLearning Strategy=Non-LLM Baseline, Model Strategy=Fine-tuning2026.01 | 74.8 | 69.3 | 38.1 | 86.2 | 67.1 | 74.34 | |
| MTL Causal LM (CLM)Learning Strategy=Multi-Task Learning (MTL), Model Strategy=Causal LM (CLM)2026.01 | 73.97 | 66.54 | 47.8 | 76.5 | 68.01 | 70.48 | |
| MTL ClassificationLearning Strategy=Multi-Task Learning (MTL), Model Strategy=Classification2026.01 | 73.9 | 66.98 | 44.79 | 81 | 66.67 | 72.4 | |
| Zero-shotLearning Strategy=Zero-shot, Model Strategy=Prompting2026.01 | 66 | 25 | 29.4 | 80.2 | 50.2 | 62.6 | |
| Few-shot ICLLearning Strategy=Few-shot, Model Strategy=In-Context Learning (ICL)2026.01 | 64.1 | 26.6 | 33.8 | 79.8 | 51.1 | 59.8 | |
| STL Causal LMLearning Strategy=Single-Task Learning (STL), Model Strategy=Causal LM2026.01 | 62.06 | 61.66 | 40.34 | 85.98 | 62.51 | 70.63 |