This paper presents our systems for the SMM4H 2026 shared task on flu-related tweetclassification across two subtasks: flu vaccination status and flu test outcome classification. For each subtask, we evaluate two approaches: fine-tuning BERTweet-large with atemporal-aware architecture, cross-validation ensembling, and regularization techniques, anda GPT-4o few-shot prompting system with similarity-based dynamic example retrieval,chain-of-thought reasoning and contrastive label ranking. Fine-tuning proves superior for theflu vaccination subtask (micro-F1: 87.90%), where sufficient and relatively balanced training datais available, while few-shot prompting performs better for the flu test subtask (micro-F1: 95.74%), where limited and heavily imbalanced training data renders fine-tuning less effective.