用合成数据增强和分语言微调，提升多语言极化检测性能。
PSK at SemEval-2026 Task 9: Multilingual Polarization Detection Using Ensemble Gemma Models with Synthetic Data Augmentation
- 为每种语言单独微调Gemma模型，结合合成数据提升泛化能力。
- 平均宏F1达0.811，3个语言排名第一，8个进入前三。
- 合成数据策略与加权集成设计有效避免过拟合，适合多语言任务。
我们提交了SemEval-2026 Task 9：多语言极化检测系统，该任务为跨22种语言的二分类任务。方法上，针对每种语言分别使用低秩适配（LoRA）微调Gemma-3（12B和27B参数）模型，并利用大语言模型（LLM）生成合成数据。采用三种合成策略（直接生成、改写、对比对构建），通过基于嵌入的去重等多阶段质量过滤流程优化数据。实验发现，在开发集上进行每语言阈值调优可带来2%至4%的F1提升，无需重新训练。同时采用12B与27B模型预测的加权集成，并根据语言选择最优策略。最终系统在所有22种语言上实现平均宏F1为0.811，排名第二，其中3个语言第一，8个语言进入前三。此外，尽管XLM-RoBERTa和Qwen3在开发集表现优异，但在测试集上出现30%至50%的F1下降，凸显泛化能力的重要性。
原文摘要 · Abstract (English)
We present our system for SemEval-2026 Task 9: Multilingual Polarization Detection, a binary classification task spanning 22 languages. Our approach fine-tunes separate Gemma~3 models (12B and 27B parameters) per language using Low-Rank Adaptation (LoRA), augmented with synthetic data generated by a large language model (LLM). We employ three synthetic data strategies (direct generation, paraphrasing, and contrastive pair creation) using GPT-4o-mini, with a multi-stage quality filtering pipeline including embedding-based deduplication. We find that per-language threshold tuning on the development set yields 2 to 4\% F1 improvements without retraining. We also use weighted ensembles of 12B and 27B model predictions with per-language strategy selection. Our final system achieves a mean macro-F1 of 0.811 across all 22 languages, ranking 2nd overall of the participating teams, with 1st place finishes in 3 languages and top-3 in 8 languages. We also find that alternative architectures (XLM-RoBERTa, Qwen3) that showed strong development set performance suffered 30 to 50\% F1 drops on the test set, highlighting the importance of generalization.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。