CSECU-DSG at SemEval-2026 Task 10: Fine-Tuning DeBERTa Transformer Model for Conspiracy Detection

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

Summary

CSECU-DSG participated in SemEval-2026 Task 10, Subtask 2, focusing on detecting conspiracy beliefs in social media posts. The team developed a transformer-based classification approach utilizing a fine-tuned DeBERTa-v3-base model to analyze Reddit comments. This method processes each post as a single input sequence. To enhance generalization and address class imbalance, the team incorporated class-weighted cross-entropy loss alongside label smoothing during the training phase. Their approach achieved competitive results, securing the ninth rank among participating teams. The findings underscore the effectiveness of fine-tuned transformer models in capturing the subtle contextual and psycholinguistic patterns inherent in conspiracy-related discourse.

Key takeaway

For Machine Learning Engineers developing misinformation detection systems, this work demonstrates that fine-tuning transformer models like DeBERTa-v3-base is a highly effective strategy. You should consider integrating techniques such as class-weighted cross-entropy loss and label smoothing to improve model robustness and performance, especially when dealing with imbalanced datasets common in social media analysis. This approach can significantly enhance your system's ability to identify subtle psycholinguistic cues.

Key insights

Fine-tuned DeBERTa models effectively detect conspiracy beliefs by capturing subtle linguistic patterns and context.

Principles

• Conspiracy detection requires understanding psycholinguistic cues.
• Contextual dependency is crucial for identifying conspiracy claims.
• Class imbalance mitigation improves model generalization.

Method

Fine-tuning a DeBERTa-v3-base model for classification, processing each post as a single input, and using class-weighted cross-entropy loss with label smoothing during training.

In practice

• Apply DeBERTa-v3-base for social media content analysis.
• Implement class-weighted loss for imbalanced datasets.
• Utilize label smoothing to enhance model generalization.

Topics

• Conspiracy Detection
• DeBERTa
• Transformer Models
• SemEval-2026
• Misinformation
• Natural Language Processing

Best for: AI Engineer, 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.
• KITE: A Tri-Modal Transformer Integrating Text, Images, and Knowledge Graphs for Fake News Detection — KITE integrates text, images, and knowledge graphs via a tri-modal transformer for enhanced fake news detection and interpretability.
• MentalMARBERT: Domain-Adaptive Pre-training and Two-Stage Fine-Tuning for Arabic Mental Health Disorders Detection — Domain-adaptive pre-training and hierarchical classification significantly improve Arabic mental health disorder detection.
• FineREX: Fine-Tuned NER-RE for Human Smuggling Knowledge Graphs — Domain-specific fine-tuning of LLMs significantly enhances knowledge graph construction quality and efficiency for complex, jargon-heavy legal texts.
• Robust Dual-Signal Fusion: Hybrid Neuro-Symbolic Gating with Compressed Chain-of-Thought Refinement for Irony Detection in Social Media Texts — Hybrid neuro-symbolic fusion with compressed Chain-of-Thought effectively overcomes LLM literalism for zero-shot irony detection.
• Comparing LLM and Fine-Tuned Model Performance on NVDRS Circumstance Extraction with Varying Prompt Complexity — LLMs, guided by a context-aware "Complexity Score" for prompt selection, significantly outperform fine-tuned models on inferentially complex, low-prevalence data extraction.

Open in AIssential →

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