@inproceedings{chowdhury-etal-2026-scope,
    title = "{SC}o{PE}: Planning for Hybrid Querying over Clinical Trial Data",
    author = "Chowdhury, Suparno  and
      Choudhury, Manan  and
      Anvekar, Tejas  and
      Khan, Muhammed  and
      Khakwani, Kaneez  and
      Sonbol, Mohamad  and
      Riaz, Irbaz  and
      Gupta, Vivek",
    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-1.76/",
    pages = "944--954",
    ISBN = "979-8-89176-434-7",
    abstract = "Systematic reviews of clinical trials require analysts to extract attributes that are rarely stored as ready-made columns. For example, the drug class of an immunotherapy named in a regimen, the additional agents combined with it, or whether a listed endpoint is a primary or secondary outcome. These attributes must be inferred from the visible content of other fields through normalization, classification, or structured extraction, and existing approaches such as direct LLM prompting, text-to-SQL, and agentic pipelines leave this reasoning implicit in a single generation step or pay a heavy execution cost for limited accuracy gains. We propose SCOPE (Structured Clinical hybrid Planning for Evidence retrieval in clinical trials), a multi-LLM planner-based framework that decomposes the task into row selection, structured planning, and execution. The planner makes the source field, reasoning rules, and output constraints explicit before answer generation, reducing ambiguity relative to direct prompting. We evaluate SCOPE on 1,500 hybrid reasoning questions over oncology clinical-trial tables against zero-shot, few-shot, chain-of-thought, TableGPT2, BlendSQL, and EHRAgent. Results show that explicit multi-LLM planning improves accuracy for reasoning-based questions while offering a stronger accuracy-efficiency tradeoff than heavier agentic baselines. Our findings position clinical trial reasoning as a distinct table understanding problem and highlight hybrid planner-based decomposition as an effective solution."
}