CCNU at SemEval-2026 Task 10: Conspiracy Marker Extraction and Detection via Multi-task Learning and LLM-based Data Augmentation

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

Summary

CCNU's system for SemEval-2026 Task 10 addresses psycholinguistic conspiracy marker extraction and conspiratorial discourse detection in Reddit comments. For Conspiracy Marker Extraction (Subtask 1), the system employs a Unified Multi-Task Sequence Labeling Framework, enabling collaborative learning across various marker types within a compact architecture. Conspiracy Detection (Subtask 2) is formulated as a sentence-level classification problem. Both subtasks benefit from data augmentation powered by large language models and ensemble inference, enhancing robustness and generalization. The system achieved strong performance, ranking 3rd on the official test set for Subtask 1 and delivering competitive results for Subtask 2.

Key takeaway

For NLP Engineers developing content moderation systems, CCNU's approach offers a robust framework. You should consider integrating multi-task sequence labeling for fine-grained text analysis, especially when identifying specific linguistic markers. LLM-based data augmentation can significantly improve model generalization and robustness, crucial for handling diverse and evolving online discourse. Implement ensemble inference to further enhance detection accuracy in real-world applications.

Key insights

Multi-task learning and LLM augmentation effectively enhance conspiracy marker extraction and detection.

Principles

• Joint modeling improves marker identification.
• LLM data augmentation boosts robustness.
• Ensemble inference enhances generalization.

Method

A Unified Multi-Task Sequence Labeling Framework jointly models multiple conspiracy markers. Sentence-level classification handles conspiracy detection. Both use LLM-based data augmentation and ensemble inference.

In practice

• Apply multi-task learning for related NLP tasks.
• Use LLMs to augment training data.
• Implement ensemble methods for robust models.

Topics

• Conspiracy Detection
• Multi-task Learning
• LLM Data Augmentation
• Sequence Labeling
• SemEval-2026 Task 10
• Reddit Discourse Analysis

Best for: Research Scientist, AI Scientist, NLP Engineer, Machine Learning 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.
• Beyond Surface Forms: A Comprehensive, Mechanism-Oriented Taxonomy of Indirect Linguistic Encoding for LLM-Based Coded Language Detection — A mechanism-oriented taxonomy of indirect linguistic encoding significantly enhances LLM-based coded language detection.
• Scaling Search Relevance: Augmenting App Store Ranking with LLM-Generated Judgments — LLM-generated textual relevance labels significantly improve App Store search ranking and conversion rates.
• Multi-View Attention Multiple-Instance Learning Enhanced by LLM Reasoning for Cognitive Distortion Detection — Combining LLMs with MIL and ELB components improves cognitive distortion detection by enhancing interpretability and reasoning.
• 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.
• Detector-Evasive LLM Paraphrasing via Constrained Policy Optimization — DEPO enables LLM paraphrasing to evade AI-text detectors while explicitly constraining semantic preservation using a Constrained Markov Decision Process.

Open in AIssential →

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