@inproceedings{aiersilan-qu-2026-agentic,
    title = "Agentic Feature Selection via {LLM} for Epileptic Seizure Detection",
    author = "Aiersilan, Aizierjiang  and
      Qu, Xiaodong",
    editor = "Demner-Fushman, Dina  and
      Ananiadou, Sophia  and
      Roberts, Kirk  and
      Tsujii, Junichi",
    booktitle = "{B}io{NLP} 2026",
    month = jul,
    year = "2026",
    address = "San Diego, California",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/ingest-acl-workshops/2026.bionlp-1.6/",
    pages = "64--74",
    ISBN = "979-8-89176-434-7",
    abstract = "Automated epileptic seizure detection from electroencephalography (EEG) signals is a clinically important task in which feature selection is typically performed using purely statistical criteria. We investigate whether a small instruction-tuned large language model (LLM) can guide iterative feature selection for binary seizure detection on the Epileptic Seizure Recognition dataset (11{\{},{\}}500 samples, 178 features). The LLM agent (Qwen2.5-1.5B-Instruct) receives five complementary statistical summaries and selects a feature subset through multi-round reasoning. The agent achieves 96.5{\textbackslash}{\%} accuracy and 0.911 F1 with 40 features, compared to 97.9{\textbackslash}{\%} accuracy and 0.946 F1 for the best full-feature baseline (SVM-RBF on 178 features). Critically, 39 of the agent{'}s 40 features coincide with the top-39 mutual-information features, and a deterministic Top-39 MI filter, evaluated by the same Random Forest classifier, attains the same 96.5{\textbackslash}{\%} accuracy and 0.911 F1. We therefore present this work as an empirical baseline: at the 1.5B-parameter scale, the LLM behaves close to a univariate MI ranker. We situate the result against the recent LLM-based feature selection literature and enumerate the ablations and multi-dataset extensions required to determine whether larger or domain-specialized LLMs add value beyond statistical filtering."
}