@inproceedings{kejriwal-etal-2026-infimobius,
    title = "Infimobius at {\#}{SMM}4{H}-{H}ea{RD} 2026: Multi-Seed {D}e{BERT}a Ensemble for Flu Vaccination and Testing Status Classification",
    author = "Kejriwal, Pradyumn  and
      Charan, Suhani Singh  and
      Sharma, Raksha  and
      Murthy, Rudra",
    editor = "Lopez-Garcia, Guillermo  and
      Gonzalez-Hernandez, Graciela",
    booktitle = "Proceedings of the 11th Social Media Mining for Health Research and Applications ({SMM}4{H}-{H}ea{RD} 2026) Workshop and Shared Tasks",
    month = jul,
    year = "2026",
    address = "San Diego, United States",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/ingest-acl-workshops/2026.smm4h-1.42/",
    pages = "260--263",
    ISBN = "979-8-89176-432-3",
    abstract = "This paper describes FluENS (Flu ENsemble System), our submission to the Social Media Mining for Health (SMM4H) 2026 Shared Task 3, which targets fine-grained classification of flu vaccination and flu testing statuses from tweets. FluENS builds on the microsoft/deberta-v2-xlarge pre-trained language model and employs a multi-seed ensemble strategy in which five models, each initialized with a different random seed and trained on the full training set, are aggregated through soft-voting over averaged softmax probabilities. We additionally incorporate balanced class weights to mitigate severe label imbalance and apply a two-stage learning rate schedule that separately controls the encoder and classification head. On the development set, FluENS achieves a macro F1 of 79.64{\%} and micro F1 of 85.56{\%} on the flu vaccination sub-task, and a macro F1 of 96.35{\%} and micro F1 of 97.04{\%} on the flu testing sub-task, substantially outperforming a roberta-base baseline across all metrics."
}