mdok-style at SemEval-2026 Task 9: Finetuning LLMs for Multilingual Polarization Detection

2026-05-04 · Source: Computation and Language · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Natural Language Processing · Depth: Advanced, quick

Summary

The mdok-style team participated in SemEval-2026 Task 9, which focuses on multilingual polarization detection across 22 languages and multiple cultural and event contexts. This task aims to identify online polarization along three axes: detection, type, and manifestation, to prevent escalation into hate speech and social fragmentation. The team addressed this challenge by finetuning mid-sized Large Language Models (LLMs) for sequence classification. They employed the QLoRA parameter-efficient finetuning technique and augmented the multilingual training data with anonymized, lower-cased, upper-cased, and homoglyphied versions to enhance detection robustness.

Key takeaway

For research scientists developing online content moderation systems, understanding the effectiveness of QLoRA with augmented multilingual data for polarization detection is critical. You should consider integrating similar data augmentation strategies and parameter-efficient finetuning techniques to improve the robustness and scalability of your models in diverse linguistic environments.

Key insights

Finetuning mid-sized LLMs with QLoRA and augmented data improves multilingual online polarization detection.

Principles

• Early polarization detection prevents online harm.
• Data augmentation enhances model robustness.

Method

Finetune mid-size LLMs for sequence classification using QLoRA, augmenting multilingual training data with anonymized, cased, and homoglyphied versions.

In practice

• Use QLoRA for efficient LLM finetuning.
• Augment text data with casing and homoglyphs.

Topics

• SemEval-2026 Task 9
• Multilingual Polarization Detection
• Large Language Models
• QLoRA Finetuning
• Sequence Classification

Best for: Research Scientist, AI Scientist, Machine Learning Engineer, NLP Engineer

Related on AIssential

• mdok-style at SemEval-2026 Task 10: Finetuning LLMs for Conspiracy Detection — Data augmentation and self-training can effectively finetune LLMs for specialized text classification tasks with limited data.
• LinguIUTics at PsyDefDetect: Iterative Imbalance-Aware Fine-tuning of Qwen3-8B for Psychological Defense Mechanism Classification — Iterative imbalance-aware fine-tuning of LLMs, combined with augmentation and post-processing, significantly improves rare-class NLP performance.
• Align Once, Benefit Multilingually: Enforcing Multilingual Consistency for LLM Safety Alignment — A plug-and-play MLC loss improves multilingual LLM safety alignment using only prompt variants, reducing resource needs.
• Detecting HIV-Related Stigma in Clinical Narratives Using Large Language Models — LLMs can effectively detect HIV-related stigma in clinical notes, with encoder models outperforming generative models in zero-shot settings.
• Improving Answer Extraction in Context-based Question Answering Systems Using LLMs — Fine-tuning pre-trained LLMs on benchmark QA datasets significantly improves answer extraction accuracy and contextual understanding in QA systems.
• Low-resource Language Discrimination Towards Chinese Dialects with Transfer learning and Data Augmentation — Transfer learning and data augmentation effectively overcome low-resource challenges in Chinese dialect discrimination.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Computation and Language.