NLP-CIMAT at SemEval-2026 Task 9: LLM-Based One-Shot and Cross-Lingual Data Augmentation for Polarization Detection

2026-06-25 · Source: Paper Index on ACL Anthology · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Natural Language Processing · Depth: Expert, medium

Summary

NLP-CIMAT participated in SemEval 2026 Task 9, "Multilingual Text Polarization." This task estimates polarization levels across languages, facing challenges from linguistic variability and limited annotated data. To address data scarcity, their pipeline combines cross-lingual translation, synthetic data augmentation via LLMs, and domain-specific pre-trained models. The approach hypothesizes that polarization signals transfer across languages without substantial semantic loss, enabling effective data augmentation. One-shot synthetic example generation proved viable for enriching training data in topic-specific scenarios. Experimental results showed high stability and competitive performance. They achieved a macro F1-score of 0.7869 for Spanish, securing 7th place. For English, they scored 0.7939, ranking 21st on the official leaderboard.

Key takeaway

For NLP Engineers building multilingual systems with scarce annotated data, consider adopting a pipeline. This pipeline integrates cross-lingual translation and LLM-based one-shot synthetic data augmentation. This approach effectively addresses data scarcity by leveraging the transferability of polarization signals across languages. You should explore one-shot generation for topic-specific data enrichment. Combine it with domain-specific pre-trained models to achieve competitive performance in tasks like multilingual polarization detection.

Key insights

LLM-based one-shot and cross-lingual data augmentation effectively addresses data scarcity for multilingual polarization detection.

Principles

• Polarization signals transfer across languages.
• One-shot generation enriches topic-specific data.
• Combine translation, LLMs, and pre-trained models.

Method

A pipeline combining cross-lingual translation, synthetic data augmentation via LLMs, and domain-specific pre-trained models to estimate multilingual text polarization levels, especially with limited annotated data.

In practice

• Use LLMs for one-shot synthetic data.
• Apply cross-lingual translation for data.
• Integrate domain-specific pre-trained models.

Topics

• Multilingual Text Polarization
• Data Augmentation
• Large Language Models
• Cross-lingual Translation
• One-Shot Learning
• SemEval 2026 Task 9

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

Related on AIssential

• mdok-style at SemEval-2026 Task 9: Finetuning LLMs for Multilingual Polarization Detection — Finetuning mid-sized LLMs with QLoRA and augmented data improves multilingual online polarization detection.
• Label-Aware Pseudo-Training Sample Generation for Text Classification — A new data augmentation method uses LLMs to subtly alter sentences, improving NLP performance with limited data.
• Understanding LLM Behavior in Multi-Target Cross-Lingual Summarization — LLMs perform cross-lingual summarization and translation jointly in later layers, not sequentially.
• Low-resource Language Discrimination Towards Chinese Dialects with Transfer learning and Data Augmentation — Transfer learning and data augmentation effectively overcome scarce annotation resources for low-resource Chinese dialect discrimination.
• English is Not All You Need: Systematically Exploring the Role of Multilinguality in LLM Post-Training — Multilingual post-training significantly improves LLM performance across languages and tasks, outperforming English-only approaches.
• Datasets for Verb Alternations across Languages: BLM Templates and Data Augmentation Strategies — This work addresses the underexplored ability of large language models (LLMs) to capture cross-sentence paradigmatic patterns, specifically verb alternations.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Paper Index on ACL Anthology.