Research map / Large language models / Language, safety and interpretability
Few-Shot Cross-Lingual Stance Detection with Sentiment-Based Pre-training
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Pre-train Pattern on sentiment-labeled Wikipedia stance examples | Pattern · 15 cross-lingual datasets, 32 shots | macro-average F1: 39.17 -> 45.09 (enWiki); 39.17 -> 45.88 (mWiki) | Helped |
| Pre-train on multilingual rather than English sentiment-labeled Wikipedia data | Pattern · 15 cross-lingual datasets, all training examples | macro-average F1: 54.82 -> 57.40 | Helped |
| Pre-train on English stance datasets rather than use Pattern without pre-training | Pattern · 15 cross-lingual datasets, 32 shots | macro-average F1: 39.17 -> 50.38 | Helped |
| Combine multi-dataset training with multilingual sentiment-based pre-training | Pattern · 15 cross-lingual datasets, all training examples | macro-average F1: 53.43 -> 62.25 | Helped |
| Train jointly on multiple datasets without pre-training | Pattern · 15 cross-lingual datasets, few-shot | macro-average F1: 39.17 -> 40.76 (32 shots); 53.43 -> 61.81 (all) | Mixed |
| Add two negative samples (ablation) | Pattern training · Few-shot stance detection | macro-average F1 | No clear effect |
| Mask some of the input tokens (ablation) | Pattern training · Few-shot stance detection | macro-average F1 | No clear effect |
| Pre-train on multilingual sentiment-based stance examples rather than English-only examples | Pattern · dast and rustance | macro F1 | Hurt |
| Train on 32 examples from English stance pre-training rather than use zero-shot inference | enstance · 15 cross-lingual datasets | macro-average F1: 35.2 -> 50.4 | Helped |
Closest papers by meaning
- MSME: A Multi-Stage Multi-Expert Framework for Zero-Shot Stance Detection (AAAI 2026 · similarity 0.87)
- Induce, Align, Predict: Zero-Shot Stance Detection via Cognitive Inductive Reasoning (AAAI 2026 · similarity 0.86)
- STEM: Unsupervised STructural EMbedding for Stance Detection (AAAI 2022 · similarity 0.86)
- Constrained Decoding for Cross-lingual Label Projection (ICLR 2024 · similarity 0.82)
- Dynamic Adapter with Semantics Disentangling for Cross-lingual Cross-modal Retrieval (AAAI 2025 · similarity 0.82)
- VERO: Verification and Zero-Shot Feedback Acquisition for Few-Shot Multimodal Aspect-Level Sentiment Classification (AAAI 2025 · similarity 0.82)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.