Elamur Research⌕
☰
home / papers / 2026 / Position: Biomedical NLP Demands Specialization, Not Generalization

Position: Biomedical NLP Demands Specialization, Not Generalization

Paper · 2026 · Toushik Wasi, Azmine

Summary

This position paper by Azmine Toushik Wasi (HeaLing 2026) questions the assumption that scaling up general-purpose models makes them robust in healthcare. The author argues that large generalist systems are poorly suited to clinical use. The reasons given are hallucinated medical claims, modality imbalance that drowns out rare biomarkers, and training data skewed toward European-ancestry patients. The paper also notes that funding concentrated on mega-models comes at the expense of rare-disease work.

It offers a theoretical framework built on three axioms: errors carry asymmetric costs, multimodal data is fragile, and interpretability is tied to usefulness. A formal "proof of criticality" based on a biomedical viability score claims that generalist models cannot structurally meet medical requirements.

As an alternative, the paper proposes a Task-First Ecosystem, a modular and clinician-in-the-loop design oriented toward equity. It supports the argument with published case studies in which narrow models beat general ones. The paper contains no new experiments. Its stated goal is to push research, funding and regulation toward specialized, auditable tools.

Key points

  • Framework: three biomedical axioms plus a viability score that combines task success probability, impact and error cost. The paper argues this score favors narrow models.
  • Task-First design principles (five):
    • putting the clinical task ahead of everything else
    • building explainability into the system from the outset
    • splitting the system into specialized components
    • repeated rounds of input from clinicians
    • ethical, equity-centered training, including federated learning and bias audits
  • Cited specialized vs. general results:
    • CNS-CLIP reached 48.84% Top-5 retrieval, compared with 18.03% for CLIP.
    • Meerkat-7B outperformed GPT-3.5 by 12% on licensing exams.
    • scELMo reached 89% accuracy on rare immune cells, 7% above scGPT.
  • Further cited results:
    • PathCLIP gained 23% over OpenAI-CLIP.
    • BiomedGPT's radiology assistant had a 3.8% report error rate, compared with 6.9% for generic models.
    • A fine-tuned BERT for sepsis triage scored 28% higher F1 than GPT-4.
    • GenePT achieved about 4× precision on ultra-rare variants.
  • Limitation: the paper is conceptual and lacks original empirical validation. It mostly relies on figures reported in other publications.

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

🎓Shahjalal University of Science and TechnologySylhet · Bangladesh2 works · 1 citation

 Fields, topics and concepts

📚 Shelves 4The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksMedicine · 241 worksHealth Informatics · 184 works

Cite

Cite as Toushik Wasi, Azmine. 2026. Position: Biomedical NLP Demands Specialization, Not Generalization. Proceedings of the 1st Workshop on Linguistic Analysis for Health (HeaLing 2026).

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.healing-1.7

@inproceedings{azmine2026position,
  title = {{Position: Biomedical NLP Demands Specialization, Not Generalization}},
  author = {Toushik Wasi, Azmine},
  booktitle = {Proceedings of the 1st Workshop on Linguistic Analysis for Health (HeaLing 2026)},
  year = {2026},
  doi = {10.18653/v1/2026.healing-1.7},
  url = {https://doi.org/10.18653/v1/2026.healing-1.7}
}

References

43 references, 5 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Gwénolé Abgrall, Andre L. Holder, Zaineb Chelly Dag-dia, Karine Zeitouni, and Xavier Monnet. 2024. Should ai models be explainable to clinicians? Critical Care, 28(1). find
  2. Julián N. Acosta, Guido J. Falcone, Pranav Rajpurkar, and Eric J. Topol. 2022. Multimodal biomedical ai. Nature Medicine, 28(9):1773–1784. find
  3. Anton Alyakin, David Kurland, Daniel Alexander Alber, Karl L. Sangwon, Danxun Li, Aristotelis Tsirigos, Eric Leuthardt, Douglas Kondziolka, and Eric Karl Oermann. 2024. Cns-clip: Transforming a neurosurgical journal into a multimodal medical model. Neurosurgery, 96(6):1227–1235. \#\# Conclusion In thi…
  4. Transactions on Computational Biology and Bioinformatics, 21(4):880–889. find
  5. Claudia C.Y. Chung, Nicole Y.T. Ng, Yvette N.C. Ng, Adrian C.Y. Lui, Jasmine L.F. Fung, Marcus C.Y. Chan, Wilfred H.S. Wong, So Lun Lee, Martin Knapp, and Brian H.Y. Chung. 2023. Socioeconomic costs of rare diseases and the risk of financial hardship: a cross-sectional study. The Lancet Regional Hea… find
  6. Jon Danielsson. 2024. When risk models hallucinate. find
  7. Emily Ernisova. 2025. Algorithmic inequities: How ai in healthcare reinforces racial disparities. https://escholarship.org/uc/item/13b8z2mq. Recent Work. Accessed: 2025-05-19. ↗
  8. FDA. 2025. Artificial intelligence and machine learning in software as a medical device. find
  9. Kiley Graim, Leslie Smith, and James Cahill. 2023. Equitable machine learning counteracts ancestral bias in precision medicine, improving outcomes for all. find
  10. Kerem Gülen. 2025. Generalist language models. Accessed: 2025-05-19. find
  11. Wolf E Hautz, Thimo Marcin, Stefanie C Hautz, Stefan K Schauber, Gert Krummrey, Martin Müller, Thomas C Sauter, Cornelia Lambrigger, David Schwappach, Mathieu Nendaz, Gregor Lindner, Simon Bosbach, Ines Griesshammer, Philipp Schönberg, Emanuel Plüss, Valerie Romann, Svenja Ravioli, Nadine Werthmülle… find
  12. Yaara Artsi, Eyal Klang, Jeremy D. Collins, Benjamin S. Glicksberg, Panagiotis Korfiatis, Girish N Nadkarni, and Vera Sorin. 2025. Large language models in radiology reporting—a systematic review of performance, limitations, and clinical implications. find
  13. Yoshua Bengio, Geoffrey Hinton, Andrew Yao, Dawn Song, Pieter Abbeel, Trevor Darrell, Yuval Noah Harari, Ya-Qin Zhang, Lan Xue, Shai Shalev-Shwartz, Gillian Hadfield, Jeff Clune, Tegan Maharaj, Frank Hutter, Atılım Güneş Baydin, Sheila McIlraith, Qiqi Gao, Ashwin Acharya, David Krueger, Anca Dragan… find
  14. Albert Q. Jiang, Alexandre Sablayrolles, Arthur Mensch, Chris Bamford, Devendra Singh Chaplot, Diego de las Casas, Florian Bressand, Gianna Lengyel, Guillaume Lample, Lucile Saulnier, Lélio Renard Lavaud, Marie-Anne Lachaux, Pierre Stock, Teven Le Scao, Thibaut Lavril, Thomas Wang, Timothée Lacroix,… find
  15. Karla Jo Helms. 2022. Millions suffer from untreatable rare diseases while medical costs skyrocket. Accessed: 2025-05-19. find
  16. Ahmad Chaddad, Qizong Lu, Jiali Li, Yousef Katib, Reem Kateb, Camel Tanougast, Ahmed Bouridane, and Ahmed Abdulkadir. 2023. Explainable, domain-adaptive, and federated artificial intelligence in medicine. IEEE/CAA Journal of Automatica Sinica, 10(4):859–876. find
  17. Hyunjae Kim, Hyeon Hwang, Jiwoo Lee, Sihyeon Park, Dain Kim, Taewhoo Lee, Chanwoong Yoon, Jiwoong Sohn, Jungwoo Park, Olga Reykhart, Thomas Fether-ston, Donghee Choi, Soo Heon Kwak, Qingyu Chen, and Jaewoo Kang. 2025a. Small language models learn enhanced reasoning skills from medical textbooks. npj…
  18. Bingyang Chen, Tao Chen, Xingjie Zeng, Weishan Zhang, Qinghua Lu, Zhaoxiang Hou, Jiehan Zhou, and Sumi Helal. 2024. Dfml: Dynamic federated meta-learning for rare disease prediction. IEEE/ACM Yubin Kim, Hyewon Jeong, Shen Chen, Shuyue Stella Li, Mingyu Lu, Kumail Alhamoud, Jimin Mun, Cristina Grau,… find
  19. Suchir Balaji, Valerie Balcom, Paul Baltescu, Haiming Bao, Mohammad Bavarian, Jeff Belgum, Irwan Bello, Jake Berdine, Gabriel Bernadett-Shapiro, Christopher Berner, Lenny Bogdonoff, Oleg Boiko, Madelaine Boyd, Anna-Luisa Brakman, Greg Brockman, Tim Brooks, Miles Brundage, Kevin Button, Trevor Cai, R… find
  20. Tianyu Liu, Tianqi Chen, Wangjie Zheng, Xiao Luo, and Hongyu Zhao. 2023. scelmo: Embeddings from language models are good learners for single-cell data analysis. find
  21. Yen-Ku Liu and Yun-Cheng Tsai. 2024. Explainable ai for trustworthy clinical decision support: A case-based reasoning system for nursing assistants. In 2024 IEEE International Conference on Big Data (BigData), page 6502–6509. IEEE. find
  22. Sheng-Chieh Lu, Christine L. Swisher, Caroline Chung, David Jaffray, and Chris Sidey-Gibbons. 2023. On the importance of interpretable machine learning predictions to inform clinical decision making in oncology. Frontiers in Oncology, 13. find
  23. Arjun Mahajan and Dylan Powell. 2025. Generalist medical ai reimbursement challenges and opportunities. npj Digital Medicine, 8(1). find
  24. Sepideh Maleki, Jan-Christian Huetter, Kangway V. Chuang, David Richmond, Gabriele Scalia, and Tommaso Biancalani. 2025. Efficient fine-tuning of single-cell foundation models enables zero-shot molecular perturbation prediction. Preprint, arXiv:2412.13478. — [arXiv:2412.13478](https://arxiv.org/abs/2412.13478) find
  25. Michael Moor, Oishi Banerjee, Zahra Shakeri Hossein Abad, Harlan M. Krumholz, Jure Leskovec, Eric J. Topol, and Pranav Rajpurkar. 2023. Foundation models for generalist medical artificial intelligence. Nature, 616(7956):259–265. find
  26. Miryam Naddaf. 2025. Ai linked to explosion of low-quality biomedical research papers. Nature. find
  27. David E. Newman-Toker, Susan M. Peterson, Shervin Badihian, Ahmed Hassoon, Najlla Nassery, Donna Parizadeh, Lisa M. Wilson, Yuanxi Jia, Rodney Omron, Saraniya Tharmarajah, Liam Guerin, Pouya B. Bastani, Elizabeth A. Fracica, Susrutha Kotwal, and Karen A. Robinson. 2022. Diagnostic Errors in the Emer… find
  28. Ayesha Siddika Nipu, K M Sajjadul Islam, and Praveen Madiraju. 2024. How reliable ai chatbots are for disease prediction from patient complaints? Preprint, arXiv:2405.13219. — [arXiv:2405.13219](https://arxiv.org/abs/2405.13219) find
  29. OpenAI, Josh Achiam, Steven Adler, Sandhini Agarwal, Lama Ahmad, Ilge Akkaya, Florencia Leoni Aleman, Diogo Almeida, Janko Altenschmidt, Sam Altman, Shyamal Anadkat, Red Avila, Igor Babuschkin, Shyam, Szymon Sidor, Eric Sigler, Maddie Simens, Jordan Sitkin, Katarina Slama, Ian Sohl, Benjamin Sokolow… — GPT-4 Technical Report
  30. Moritani, Sevcan Turk, Joonsang Lee, Snehal Prabhudesai, Fanny Morón, Jacob Mandel, Konstantinos Kamnitsas, Ben Glocker, Luke V. M. Dixon, Matthew Williams, Peter Zampakis, Vasileios Panagiotopoulos, Panagiotis Tsiganos, Sotiris Alexiou, Ilias Haliassos, Evangelia I. Zacharaki, Konstantinos Moustaka… find
  31. Sarthak Pati, Ujjwal Baid, Brandon Edwards, Micah Sheller, Shih-Han Wang, G. Anthony Reina, Patrick Foley, Alexey Gruzdev, Deepthi Karkada, Christos Davatzikos, Chiharu Sako, Satyam Ghodasara, Michel Bilello, Suyash Mohan, Philipp Vollmuth, Gianluca Brugnara, Chandrakanth J. Preetha, Felix Sahm, Kla… find
  32. Rajanya Roy, Rohit Bansal, David Tegay, and Sara Benolkin. 2023. How rare diseases add up, making them a high public health priority. find
  33. Emre Sezgin and Ahmet Baki Kocaballi. 2025. Era of generalist conversational artificial intelligence to support public health communications. Journal of Medical Internet Research, 27:e69007. find
  34. Becca Roelofs, Sid Lall, Amit Marathe, Betty Chan, Santiago Ontanon, Luheng He, Denis Teplyashin, Jonathan Lai, Phil Crone, Bogdan Damoc, Lewis Ho, Sebastian Riedel, Karel Lenc, Chih-Kuan Yeh, Aakanksha Chowdhery, Yang Xu, Mehran Kazemi, Ehsan Amid, Anastasia Petrushkina, Kevin Swersky, Ali Khodaei,… find
  35. Benjamin D. Simon, Kutsev Bengisu Ozyoruk, David G. Gelikman, Stephanie A. Harmon, and Barış Türkbey. 2024. The future of multimodal artificial intelligence models for integrating imaging and clinical metadata: a narrative review. Diagnostic and Interventional Radiology. find
  36. Karan Singhal, Shekoofeh Azizi, Tao Tu, S. Sara Mahdavi, Jason Wei, Hyung Won Chung, Nathan Scales, Ajay Tanwani, Heather Cole-Lewis, Stephen Pfohl, Perry Payne, Martin Seneviratne, Paul Gamble, Chris Kelly, Abubakr Babiker, Nathanael Schärli, Aakanksha Chowdhery, Philip Mansfield, Dina Demner-Fushm… — Large Language Models Encode Clinical Knowledge
  37. Tulsi Suchak, Anietie E. Aliu, Charlie Harrison, Reyer Zwiggelaar, Nophar Geifman, and Matt Spick. 2025. Explosion of formulaic research articles, including inappropriate study designs and false discoveries, based on the nhanes us national health database. PLOS Biology, 23(5):e3003152. find
  38. Gemini Team, Petko Georgiev, Ving Ian Lei, Ryan Burnell, Libin Bai, Anmol Gulati, Garrett Tanzer, Damien Vincent, Zhufeng Pan, Shibo Wang, Soroosh Mariooryad, Yifan Ding, Xinyang Geng, Fred Alcober, Roy Frostig, Mark Omernick, Lexi Walker, Cosmin Paduraru, Christina Sorokin, Andrea Tacchetti, Colin… — Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context
  39. Han, Woosuk Kwon, Xi Chen, Yinlam Chow, Yuvein Zhu, Zichuan Wei, Zoltan Egyed, Victor Cotruta, Minh Giang, Phoebe Kirk, Anand Rao, Kat Black, Nabila Babar, Jessica Lo, Erica Moreira, Luiz Gustavo Martins, Omar Sanseviero, Lucas Gonzalez, Zach Gleicher, Tris Warkentin, Vahab Mirrokni, Evan Senter, El… find
  40. Gemma Team, Aishwarya Kamath, Johan Ferret, Shreya Pathak, Nino Vieillard, Ramona Merhej, Sarah Perrin, Tatiana Matejovicova, Alexandre Ramé, Morgane Rivière, Louis Rouillard, Thomas Mesnard, Geoffrey Cideron, Jean bastien Grill, Sabela Ramos, Edouard Yvinec, Michelle Casbon, Etienne Pot, Ivo Penche… find
  41. Kai Zhang, Rong Zhou, Eashan Adhikarla, Zhiling Yan, Yixin Liu, Jun Yu, Zhengliang Liu, Xun Chen, Brian D. Davison, Hui Ren, Jing Huang, Chen Chen, Yuyin Zhou, Sunyang Fu, Wei Liu, Tianming Liu, Xiang Li, Yong Chen, Lifang He, James Zou, Quanzheng Li, Hongfang Liu, and Lichao Sun. 2024. A generalist… find
  42. Sheng Zhang, Yanbo Xu, Naoto Usuyama, Hanwen Xu, Jaspreet Bagga, Robert Tinn, Sam Preston, Rajesh Rao, Mu Wei, Naveen Valluri, Cliff Wong, Andrea Tupini, Yu Wang, Matt Mazzola, Swadheen Shukla, Lars Liden, Jianfeng Gao, Angela Crabtree, Brian Piening, Carlo Bifulco, Matthew P. Lungren, Tristan Nauma… find
  43. Sunyi Zheng, Xiaonan Cui, Yuxuan Sun, Jingxiong Li, Honglin Li, Yunlong Zhang, Pingyi Chen, Xueping Jing, Zhaoxiang Ye, and Lin Yang. 2024. Benchmarking pathclip for pathology image analysis. Journal of Imaging Informatics in Medicine, 38(1):422–438. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/wasi-2026-position-biomedical-nlp-demands-specialization/