ttda704 at SemEval-2026 Task 6: Structured Chain-of-Thought Prompting for Political Evasion Detection

The paper presents a system for the SemEval-2026 Task 6 focused on detecting political evasion in U.S. presidential interviews, utilizing two approaches: Parameter-Efficient Fine-Tuning of Qwen3 models (4B-32B) with QLoRA and structured Chain-of-Thought (CoT) prompting with DeepSeek-V3.2 and Grok-4-Fast. The results show that Grok-4-Fast, enhanced with hierarchical CoT prompting, achieved a Macro F1 score of 0.5147 on a 9-class evasion task and 0.7979 on a 3-class clarity task, outperforming the fine-tuning baseline. This work highlights the importance of effective prompt design and the advantages of enabling extended reasoning for improved detection of evasive language, providing valuable insights for practitioners in model training and prompt engineering.