Elamur Research⌕
☰
home / papers / 2026 / Scaling is Not All You Need: Clinical-Oriented Reinforcement Learning Makes Parameter-Efficient Clinical Reasoning

Scaling is Not All You Need: Clinical-Oriented Reinforcement Learning Makes Parameter-Efficient Clinical Reasoning

Paper · 2026 · Chi Liu, Yan Shu, Mengzhuo Chen et al.

Summary

Strong medical reasoning in LLMs usually depends on very large models and costly hand-annotated chain-of-thought data. The authors argue that a carefully designed training pipeline can matter more than model size. They introduce CORL (Clinical-Oriented Reinforcement Learning), presented as the first fully open end-to-end RL pipeline for clinical reasoning. It has three parts:

  • RODS data strategy. Public medical QA sets are mixed with questions synthesized from a knowledge graph built from Wikipedia. The synthesis samples multi-hop subgraphs, masks parts of them, deliberately favours rare diseases and drugs, and adds plausible but wrong distractors.
  • CoT cold start. Reasoning traces are distilled from teacher models, filtered with pass@k selection, and checked for verifiability.
  • Two-stage RLVR with GRPO. The first stage trains on moderate-difficulty items. The second mines hard cases the model still fails.

The resulting Fleming-R1 models perform strongly on medical benchmarks. The 7B variant beats peers of its size and some 72B models on MedXpertQA. The 32B variant essentially ties GPT-4o. This suggests compact, openly released models can deliver expert-level clinical QA more cheaply. Evaluation, however, is limited to closed-ended QA.

Key points

  • Training data: MedQA, MedMCQA, CMExam and PubMedQA, plus synthetic items from a Wikipedia-derived medical knowledge graph with over 100,000 entities.
  • RL data mixes: Stage 1 balances hard and simple samples 1:1. Stage 2 uses a 1:2 ratio of right to wrong samples. Rewards cover only answer correctness and format.
  • MedXpertQA (4,460 questions, 17 specialties), 7B scale: Fleming-R1-7B scores 19.14%, compared with 11.96% for Qwen2.5-7B-Instruct and 18.90% for Qwen2.5-72B.
  • MedXpertQA, 32B scale: Fleming-R1-32B scores 30.33%, against 30.37% for GPT-4o, 26.04% for Qwen3-32B and 23.39% for Baichuan-M2-32B.
  • Ablation: across benchmarks, two-stage RLVR raises the average score from 55.4 to 63.4 at 7B and from 72.9 to 75.4 at 32B. Code, data and models are released on GitHub (UbiquantAI/Fleming-R1) and Hugging Face (IQuestLab).

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

 Fields, topics and concepts

📚 Shelves 5The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksComputer Vision and Pattern Recognition · 2,869 worksMedicine · 241 worksHealth Informatics · 184 works

🧭 Topics 1Research subjects and languages the work is about.Reasoning · 1,426 works

Cite

Cite as Chi Liu, Yan Shu, Mengzhuo Chen, Hongming Piao, Zhijian Duan, Derek Li and Bryan Dai. 2026. Scaling is Not All You Need: Clinical-Oriented Reinforcement Learning Makes Parameter-Efficient Clinical Reasoning. Findings of the Association for Computational Linguistics: ACL 2026.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.findings-acl.741

@inproceedings{liu2026scaling,
  title = {{Scaling is Not All You Need: Clinical-Oriented Reinforcement Learning Makes Parameter-Efficient Clinical Reasoning}},
  author = {Chi Liu, Yan Shu, Mengzhuo Chen, Hongming Piao, Zhijian Duan, Derek Li and Bryan Dai},
  booktitle = {Findings of the Association for Computational Linguistics: ACL 2026},
  year = {2026},
  doi = {10.18653/v1/2026.findings-acl.741},
  url = {https://doi.org/10.18653/v1/2026.findings-acl.741}
}

References

36 references, 15 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Manar Aljohani, Jun Hou, Sindhura Kommu, and Xuan Wang. 2025. A comprehensive survey on the trustworthiness of large language models in healthcare. arXiv preprint arXiv:2502.15871. find
  2. Yasmeen Alufaisan, Laura R Marusich, Jonathan Z Bakdash, Yan Zhou, and Murat Kantarcioglu. 2021. Does explainable artificial intelligence improve human decision-making? In Proceedings of the AAAI Conference on Artificial Intelligence, volume 35, pages 6618–6626. find
  3. Anna Arias-Duart, Pablo Agustin Martin-Torres, Daniel Hinjos, Pablo Bernabeu-Perez, Lucia Urcelay Ganzabal, Marta Gonzalez Mallo, Ashwin Kumar Gururajan, Enrique Lopez-Cuena, Sergio Alvarez-Napagao, and Dario Garcia-Gasulla. 2025. Automatic evaluation of healthcare LLMs beyond question-answering. In… find
  4. Junying Chen, Zhenyang Cai, Ke Ji, Xidong Wang, Wanlong Liu, Rongsheng Wang, and Benyou Wang. 2025. Towards medical complex reasoning with LLMs through medical verifiable problems. In Findings of the Association for Computational Linguistics: ACL 2025, pages 14552–14573, Vienna, Austria. Association… find
  5. Kristof Coussement, Mohammad Zoynul Abedin, Mathias Kraus, Sebastián Maldonado, and Kazim Topuz. 2024. Explainable ai for enhanced decision-making. Decision Support Systems, 184:114276. find
  6. Chengfeng Dou, Chong Liu, Fan Yang, Fei Li, Jiyuan Jia, Mingyang Chen, Qiang Ju, Shuai Wang, Shunya Dang, Tianpeng Li, and 1 others. 2025. Baichuanm2: Scaling medical capability with large verifier system. arXiv preprint arXiv:2509.02208.
  7. Daya Guo, Dejian Yang, Haowei Zhang, Junxiao Song, Ruoyu Zhang, Runxin Xu, Qihao Zhu, Shirong Ma, Peiyi Wang, Xiao Bi, and 1 others. 2025. Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning. arXiv preprint arXiv:2501.12948.
  8. Paul Hager, Friederike Jungmann, Robbie Holland, Kunal Bhagat, Inga Hubrecht, Manuel Knauer, Jakob Vielhauer, Marcus Makowski, Rickmer Braren, Georgios Kaissis, and 1 others. 2024. Evaluation and mitigation of the limitations of large language models in clinical decision-making. Nature medicine, 30(… find
  9. Di Jin, Eileen Pan, Nassim Oufattole, Wei-Hung Weng, Hanyi Fang, and Peter Szolovits. 2021. What disease does this patient have? a large-scale open domain question answering dataset from medical exams. Applied Sciences, 11(14):6421.
  10. 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…
  11. Girish Joseph, Neena Bhatti, Rithik Mittal, and Arun Bhatti. 2025. Current application and future prospects of artificial intelligence in healthcare and medical education: A review of literature. Cureus, 17(1):e77313. find
  12. Takeshi Kojima, Shixiang Shane Gu, Machel Reid, Yutaka Matsuo, and Yusuke Iwasawa. 2022. Large language models are zero-shot reasoners. Advances in neural information processing systems, 35:22199– 22213.
  13. Zeljko Kraljevic, Anthony Shek, Daniel Bean, Rebecca Bendayan, James Teo, and Richard Dobson. 2021. Medgpt: Medical concept prediction from clinical narratives. arXiv preprint arXiv:2107.03134. — [arXiv:2107.03134](https://arxiv.org/abs/2107.03134) find
  14. Yuxiang Lai, Jike Zhong, Ming Li, Shitian Zhao, and Xiaofeng Yang. 2025. Med-r1: Reinforcement learning for generalizable medical reasoning in vision-language models. arXiv preprint arXiv:2503.13939. — [arXiv:2503.13939](https://arxiv.org/abs/2503.13939) find
  15. Cheng Li, Mengzhuo Chen, Jindong Wang, Sunayana Sitaram, and Xing Xie. 2024. Culturellm: Incorporating cultural differences into large language models. Advances in Neural Information Processing Systems, 37:84799–84838.
  16. Che Liu, Haozhe Wang, Jiazhen Pan, Zhongwei Wan, Yong Dai, Fangzhen Lin, Wenjia Bai, Daniel Rueckert, and Rossella Arcucci. 2025a. Beyond distillation: Pushing the limits of medical llm reasoning with minimalist rule-based rl. arXiv preprint arXiv:2505.17952.
  17. Fenglin Liu, Hongjian Zhou, Boyang Gu, Xinyu Zou, Jinfa Huang, Jinge Wu, Yiru Li, Sam S. Chen, Yining Hua, Peilin Zhou, Junling Liu, Chengfeng Mao, Chenyu You, Xian Wu, Yefeng Zheng, Lei Clifton, Zheng Li, Jiebo Luo, and David A. Clifton. 2025b. Application of large language models in medicine. Natu… find
  18. Jiaxiang Liu, Yuan Wang, Jiawei Du, Joey Tianyi Zhou, and Zuozhu Liu. 2024. MedCoT: Medical chain of thought via hierarchical expert. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 17371–17389, Miami, Florida, USA. Association for Computational Lingu… find
  19. Junling Liu, Peilin Zhou, Yining Hua, Dading Chong, Zhongyu Tian, Andrew Liu, Helin Wang, Chenyu You, Zhenhua Guo, Lei Zhu, and 1 others. 2023. Benchmarking large language models on cmexam-a comprehensive chinese medical exam dataset. Advances in Neural Information Processing Systems, 36:52430–52452… find
  20. 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
  21. 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.
  22. Long Ouyang, Jeffrey Wu, Xu Jiang, Diogo Almeida, Carroll Wainwright, Pamela Mishkin, Chong Zhang, Sandhini Agarwal, Katarina Slama, Alex Ray, and 1 others. 2022. Training language models to follow instructions with human feedback. Advances in neural information processing systems, 35:27730–27744.
  23. Ankit Pal, Logesh Kumar Umapathi, and Malaikannan Sankarasubbu. 2022. Medmcqa: A large-scale multi-subject multi-choice dataset for medical domain question answering. In Conference on health, inference, and learning, pages 248–260. PMLR.
  24. Mubashar Raza, Zarmina Jahangir, Muhammad Bilal Riaz, Muhammad Jasim Saeed, and Muhammad Awais Sattar. 2025. Industrial applications of large language models. Scientific Reports, 15(1):13755. find
  25. John Schulman, Filip Wolski, Prafulla Dhariwal, Alec Radford, and Oleg Klimov. 2017. Proximal policy optimization algorithms. arXiv preprint arXiv:1707.06347. find
  26. Zhihong Shao, Peiyi Wang, Qihao Zhu, Runxin Xu, Junxiao Song, Xiao Bi, Haowei Zhang, Mingchuan Zhang, YK Li, Yang Wu, and 1 others. 2024. Deepseekmath: Pushing the limits of mathematical reasoning in open language models. arXiv preprint arXiv:2402.03300.
  27. 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
  28. Karan Singhal, Tao Tu, Juraj Gottweis, Rory Sayres, Ellery Wulczyn, Mohamed Amin, Le Hou, Kevin Clark, Stephen R Pfohl, Heather Cole-Lewis, and 1 others. 2025. Toward expert-level medical question answering with large language models. Nature Medicine, 31(3):943–950. find
  29. Qiyang Sun, Alican Akman, and Björn W Schuller. 2025. Explainable artificial intelligence for medical applications: A review. ACM Transactions on Computing for Healthcare, 6(2):1–31. find
  30. Yubo Wang, Xueguang Ma, Ge Zhang, Yuansheng Ni, Abhranil Chandra, Shiguang Guo, Weiming Ren, Aaran Arulraj, Xuan He, Ziyan Jiang, and 1 others. 2024. Mmlu-pro: A more robust and challenging multi-task language understanding benchmark. Advances in Neural Information Processing Systems, 37:95266–95290…
  31. Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Fei Xia, Ed Chi, Quoc V Le, Denny Zhou, and 1 others. 2022. Chain-of-thought prompting elicits reasoning in large language models. Advances in neural information processing systems, 35:24824– 24837.
  32. An Yang, Anfeng Li, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chang Gao, Chengen Huang, Chenxu Lv, and 1 others. 2025. Qwen3 technical report. arXiv preprint arXiv:2505.09388.
  33. Sheng Zhang, Qianchu Liu, Guanghui Qin, Tristan Naumann, and Hoifung Poon. 2025. Med-rlvr: Emerging medical reasoning from a 3b base model via reinforcement learning. arXiv preprint arXiv:2502.19655. find
  34. Yuxin Zuo, Shang Qu, Yifei Li, Zhangren Chen, Xuekai Zhu, Ermo Hua, Kaiyan Zhang, Ning Ding, and Bowen Zhou. 2025. Medxpertqa: Benchmarking expert-level medical reasoning and understanding. arXiv preprint arXiv:2501.18362. find
  35. Qwen Team. 2024. Qwen2.5: A party of foundation models. find
  36. A Dandan Wang and Shiqing Zhang. 2024. Large language models in medical and healthcare fields: applications, advances, and challenges. Artificial Intelligence Review, 57(11):299. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/liu-2026-scaling-is-not-all-you/