We describe our system for Subtask 4 (Evidence Alignment) of the ArchEHR-QA 2026 shared task, which requires aligning each sentence of a clinician-authored answer to the supporting sentence(s) in a clinical note excerpt derived from MIMIC. The task is challenging due to many-to-many alignment structure, answer sentences with no note support, and the semantic gap between clinical note language and answer paraphrases. We explore two approaches: few-shot chain-of-thought prompting with Qwen2.5-7B-Instruct and LoRA fine-tuning of a cross-encoder with combined InfoNCE and BCE loss. Our best system achieves a micro F1 of 67.93 on the test set.

We present a retrieval-augmented system for the ArchEHR-QA 2025 shared task, which focuses on generating concise, medically accurate answers to clinical questions based on a patient’s electronic health record (EHR). A key challenge is following a strict cita- tion format that references relevant sentence IDs. To improve retrieval, we fine-tuned an all-MiniLM-L6-v2 embedding model using contrastive learning on over 2,300 question–sentence triplets, with DoRA for efficient adaptation. Sentences were selected using cosine similarity thresholds and passed into a quantized Mistral-7B-Instruct model along with a structured prompt. Our system achieved similar relevance to the baseline but lower overall performance (19.3 vs. 30.7), due to issues with citation formatting and generation quality. We discuss limitations such as threshold tuning, prompt-following ability, and model size, and suggest future directions for improving structured biomedical QA.