NIT-Agartala-NLP-Team at SemEval-2026 Task 9: A Weighted Soft-Voting Ensemble Framework of Fine-Tuned LLMs for Binary and Multi-Label Polarization Detection

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

Summary

The NIT-Agartala-NLP-Team submitted a weighted soft-voting ensemble framework of fine-tuned large language models (LLMs) to SemEval-2026 Task 9 for polarization detection in textual data. This task involved two subtasks: binary classification to distinguish polarized from non-polarized content, and multi-label classification to identify specific polarization types. Their framework combined probabilistic outputs from individual LLMs using weighted averaging, aiming to leverage complementary strengths and enhance overall performance. The system achieved a test macro F1-score of 78.6 (26th out of 44 teams) in Subtask 1 and 46.0 (18th out of 29 teams) in Subtask 2, demonstrating the utility of ensemble methods for complex text classification challenges.

Key takeaway

For NLP Engineers developing robust text classification systems, this work highlights the effectiveness of weighted soft-voting ensembles. You should consider integrating multiple fine-tuned LLMs and combining their probabilistic outputs to improve performance on complex tasks like polarization detection. This approach can leverage diverse model strengths, potentially yielding better results than single models, especially for multi-label challenges.

Key insights

Weighted soft-voting ensembles of fine-tuned LLMs enhance polarization detection by combining complementary model strengths.

Principles

• Combine probabilistic outputs from diverse models.
• Leverage complementary strengths of individual LLMs.

Method

Integrate multiple fine-tuned LLMs, then combine their probabilistic outputs using weighted averaging to enhance overall performance.

In practice

• Binary classification for polarized vs. non-polarized content.
• Multi-label classification for specific polarization types.

Topics

• SemEval-2026
• Polarization Detection
• Ensemble Learning
• Large Language Models
• Text Classification
• Weighted Soft-Voting

Best for: Research Scientist, AI Scientist, NLP Engineer, Machine Learning 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.
• TeamHerald@CHIPSAL 2026: Hate Speech Detection and Sentiment Analysis of Nepali Memes using Transformer-based Architectures and Ensemble Learning — Ensemble strategies for meme analysis perform differently for binary vs. multi-class tasks, requiring task-specific aggregation.
• Enhancing SDG-Text Classification with Combinatorial Fusion Analysis and Generative AI — Combinatorial Fusion Analysis (CFA) significantly enhances SDG text classification by combining diverse models and synthetic data.
• Consistency Analysis of Sentiment Predictions using Syntactic & Semantic Context Assessment Summarization (SSAS) — SSAS improves LLM sentiment prediction consistency by pre-processing noisy data into structured, high-signal prompts.
• Comparing BERT Sentence-Pair Classification and Few-Shot LLM Prompting for Detecting Threat and Solution Framing in German Climate News — Fine-tuned BERT models outperform few-shot LLMs for German climate news framing detection, especially with sentence context.
• From Logistic Regression to GPT-2: Building a Complete Spam Detection & Sentiment Analysis Pipeline — For imbalanced text classification, prioritize F1-score and Precision-Recall AUC over accuracy and ROC-AUC.

Open in AIssential →

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