Team YTY at SemEval 2026 task 12: Option-Aware Retrieval and Cross-Encoder Reasoning Framework for Abductive Event Reasoning

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

Summary

Team YTY developed a unified system for SemEval-2026 Task 9, focusing on multilingual polarization detection. This task encompasses binary polarization detection, multi-label target type classification, and multi-label manifestation identification across various languages and events, notably addressing severe class imbalance. Their approach integrates targeted data augmentation for low-frequency labels, merged multitask fine-tuning for Subtask 2 and Subtask 3, and model fusion to enhance cross-lingual stability. Subtask 1 predictions are generated through calibrated inference from the multi-label head. On the development set, multitask training consistently surpassed single-task methods, with model fusion providing further improvements, particularly for rare labels. Remaining challenges include implicit polarization and partial-label uncertainty.

Key takeaway

For NLP Engineers developing multilingual text classification systems, especially those facing severe class imbalance and multiple related subtasks, you should prioritize integrating targeted data augmentation for rare labels. Consider a merged multitask fine-tuning approach for interdependent subtasks and implement model fusion to significantly improve cross-lingual stability and overall performance on challenging datasets. This strategy can yield substantial gains over single-task methods.

Key insights

Multitask fine-tuning and model fusion enhance multilingual polarization detection, especially for imbalanced rare labels.

Principles

• Targeted data augmentation improves low-frequency label performance.
• Multitask training consistently outperforms single-task variants.
• Model fusion enhances cross-lingual stability and rare label gains.

Method

The system combines targeted data augmentation, merged multitask fine-tuning of Subtask 2 and 3, and model fusion. Subtask 1 predictions use calibrated inference from the multi-label head.

In practice

• Apply data augmentation for imbalanced classification tasks.
• Consider multitask learning for related subtasks.
• Implement model fusion for cross-lingual robustness.

Topics

• Multilingual Polarization Detection
• SemEval-2026 Task 9
• Data Augmentation
• Multitask Learning
• Model Fusion
• Class Imbalance

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

Related on AIssential

• Enhancing Multilingual Reasoning via Steerable Model Merging — Steerable Model Merging adaptively modulates source model contributions to resolve conflicts and enhance multilingual reasoning.
• SURGELLM: Rethinking Multi-Task Evaluation through Task-Aware Feature Gating with Class-Balanced Normalization — SURGELLM enhances multi-task NLP performance by integrating task-aware feature gating and class-balanced normalization within a unified transformer framework.
• Human Label Variation as Stable Signal: Learning Annotator-Specific Explanation Behavior via Cross-Annotator Preference Optimization — Large language models can learn and reproduce individual annotator explanation styles from human label variation.
• 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.
• DREAM: Extending Vision-Language Models with Dual-Objective Encoding for Cross-Modal Retrieval — Dual-objective encoding and hierarchical attention significantly enhance cross-modal video retrieval performance.
• A Multitask Transformer for Offensive Language Detection and Target Identification in HateBR — Multitask learning improves offensive language detection consistency but may reduce fine-grained target attribution.

Open in AIssential →

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