Elamur Research⌕
☰
home / papers / 2026 / TAMU-NLP at ArchEHR-QA 2026: Grounded Clinical QA with Evidence Identification and Intent-Aware Answer Generation

TAMU-NLP at ArchEHR-QA 2026: Grounded Clinical QA with Evidence Identification and Intent-Aware Answer Generation

Paper · 2026 · Xinqi Su, Rongrong Wang, Sunyang Fu et al.

Summary

This paper describes the TAMU-NLP system for the ArchEHR-QA 2026 shared task. The task asks systems to answer patient questions using evidence from that patient's own electronic health record. The team took part in two of the subtasks, finding supporting evidence (Subtask 2) and producing answers (Subtask 3).

Annotated data was scarce, with only 20 labelled development cases. To get training data, the authors had an LLM label evidence sentences in 100 further cases, using the clinician answers as a guide. The labels were reduced to two classes: essential versus everything else.

  • Evidence identification: they compared cross-encoder learning-to-rank models against LLMs. The LLM approach works in two stages. First, few-shot prompting labels each sentence. Second, a self-reflection pass reviews and corrects those labels, followed by ensemble voting to make predictions more stable.
  • Answer generation: an LLM first decides what kind of question is being asked and how risky it is. A few-shot prompt then writes a short answer drawn strictly from the clinical note.

The system placed 5th in Subtask 2 and 2nd in Subtask 3. This suggests that pairing evidence-focused reasoning with LLM generation is a practical route to grounded clinical QA when data is limited.

Key points

  • Data: the ArchEHR-QA 2026 dataset uses clinical notes from MIMIC.
    • The development set (Cases 1–20) contains 121 essential and 307 not-essential sentences, so the classes are imbalanced.
    • Cases 21–120 received LLM-generated labels and served as training data.
    • The test set covers Cases 121–167 (47 cases).
  • Ranking models: the cross-encoder takes only the clinician question as the query; adding the patient narrative helped little. Training objectives included pointwise (weighted BCE), pairwise with hard negative mining, listwise, and hybrid losses.
  • Intent classification: each question is assigned a primary intent from 12 categories (for example medication_safety, prognosis_severity, activity_clearance). The model may also give a secondary intent and a low/moderate/high risk level.
  • Answer constraints: answers are limited to 75 words in 2–4 sentences. The prompt asks the model to reuse wording from the note and to cover the key clinical concepts.
  • Evaluation and results: Subtask 3 was scored with BLEU, ROUGE, BERTScore, SARI, AlignScore and MEDCON. The official leaderboard placed the system 5th in Subtask 2 and 2nd in Subtask 3.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

🎓Texas A&M UniversityCollege Station · United States26 works · 1,871 citations🎓The University of Texas Health Science Center at HoustonHouston · United States5 works · 36 citations

 Fields, topics and concepts

📚 Shelves 6The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksArts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 worksMedicine · 241 worksHealth Informatics · 184 works

🧭 Topics 1Research subjects and languages the work is about.Question answering · 441 works

Cite

Cite as Xinqi Su, Rongrong Wang, Sunyang Fu, Hongfang Liu and Ruihong Huang. 2026. TAMU-NLP at ArchEHR-QA 2026: Grounded Clinical QA with Evidence Identification and Intent-Aware Answer Generation. Proceedings of the Third Workshop on Patient-Oriented Language Processing (CL4Health) @ LREC 2026.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.63317/55c9e7av4wry

@inproceedings{su2026tamu,
  title = {{TAMU-NLP at ArchEHR-QA 2026: Grounded Clinical QA with Evidence Identification and Intent-Aware Answer Generation}},
  author = {Xinqi Su, Rongrong Wang, Sunyang Fu, Hongfang Liu and Ruihong Huang},
  booktitle = {Proceedings of the Third Workshop on Patient-Oriented Language Processing (CL4Health) @ LREC 2026},
  year = {2026},
  doi = {10.63317/55c9e7av4wry},
  url = {https://doi.org/10.63317/55c9e7av4wry}
}

References

22 references, 9 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Asma Ben Abacha and Dina Demner-Fushman. 2019. A question-entailment approach to question answering. BMC Bioinformatics, 20(1):511. find
  2. Asma Ben Abacha and Pierre Zweigenbaum. 2015. Means: A medical question-answering system combining nlp techniques and semantic web technologies. Information Processing & Management, 51(5):570–594. find
  3. Christopher Burges. 2010. From ranknet to lambdarank to lambdamart: An overview. Learning, 11:81. find
  4. Zhe Cao, Tao Qin, Tie-Yan Liu, Ming-Feng Tsai, and Hang Li. 2007. Learning to rank: From pairwise approach to listwise approach. In Proceedings of ICML, pages 129–136. find
  5. Souradip Chakraborty, Ekaba Bisong, Shweta Bhatt, Thomas Wagner, Riley Elliott, and Francesco Mosconi. 2020. Biomedbert: A pre-trained biomedical language model for qa and ir. In Proceedings of COLING, pages 669–679. find
  6. Robert M Cronin, SE Davis, JA Shenson, Q Chen, ST Rosenbloom, and GP Jackson. 2015. Growth of secure messaging through a patient portal as a form of outpatient interaction across clinical specialties. Applied Clinical Informatics, 6(02):288– 304. find
  7. Dina Demner-Fushman and Jimmy Lin. 2007. Answering clinical questions with knowledge-based and statistical techniques. Computational Linguistics, 33(1):63–103. find
  8. Taya Irizarry, Annette DeVito Dabbs, and Christine Curran. 2015. Patient portals and patient engagement: a state of the science review. Journal of Medical Internet Research, 17(6):e148. find
  9. Qiao Jin, Bhuwan Dhingra, Zhengping Liu, William Cohen, and Xinghua Lu. 2019. Pubmedqa: A dataset for biomedical research question answering. In Proceedings of the 2019 conference on empirical methods in natural language processing and the 9th international joint conference on natural language proce…
  10. Jinhyuk Lee, Wonjin Yoon, Sungdong Kim, Donghyeon Kim, Sunkyu Kim, Chan Ho So, and Jaewoo Kang. 2020. Biobert: a pre-trained biomedical language representation model for biomedical text mining. Bioinformatics, 36(4):1234–1240.
  11. Tsung-Yi Lin, Priya Goyal, Ross Girshick, Kaiming He, and Piotr Dollár. 2017. Focal loss for dense object detection. In Proceedings of ICCV, pages 2980–2988.
  12. Large language models encode clinical knowledge. Nature, 620(7972):172–180.
  13. Sarvesh Soni and Dina Demner-Fushman. 2026a. A dataset for addressing patient’s information needs related to clinical course of hospitalization. Scientific Data. find
  14. Sarvesh Soni and Dina Demner-Fushman. 2026b. Overview of the archehr-qa 2026 shared task on grounded question answering from electronic health records. In Proceedings of the Third Workshop on Patient-Oriented Language Processing (CL4Health), Palma, Mallorca (Spain). ELRA. find
  15. Wei Xu, Courtney Napoles, Ellie Pavlick, Quanze Chen, and Chris Callison-Burch. 2016. Optimizing statistical machine translation for text simplification. Transactions of the Association for Computational Linguistics, 4:401–415.
  16. Wen-wai Yim, Yujuan Fu, Asma Ben Abacha, Neal Snider, Thomas Lin, and Meliha Yetisgen. 2023. Aci-bench: a novel ambient clinical intelligence dataset for benchmarking automatic visit note generation. Scientific data, 10(1):586. find
  17. Rodrigo Nogueira and Kyunghyun Cho. 2019. Passage re-ranking with bert. arXiv preprint arXiv:1901.04085.
  18. Harsha Nori, Nicholas King, Scott Mayer McKinney, Dean Carignan, and Eric Horvitz. 2023. Capabilities of gpt-4 on medical challenge problems. arXiv preprint arXiv:2303.13375.
  19. Anusri Pampari, Preethi Raghavan, Jennifer Liang, and Jian Peng. 2018. emrqa: A large corpus for question answering on electronic medical records. In Proceedings of the 2018 conference on empirical methods in natural language processing, pages 2357–2368.
  20. Nils Reimers and Iryna Gurevych. 2019. Sentence-bert: Sentence embeddings using siamese bert-networks. In Proceedings of EMNLP-IJCNLP, pages 3982–3992.
  21. Anna Rogers, Jordan Boyd-Graber, and Naoaki Okazaki. 2023. Proceedings of the 61st annual meeting of the association for computational linguistics (volume 1: Long papers). In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). find
  22. Karan Singhal, Shekoofeh Azizi, Tao Tu, Sara Mahdavi, Jason Wei, Hyung Won Chung, et al. 2023. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/su-2026-tamu-nlp-archehr-qa-2026/