Elamur Research⌕
☰
home / papers / 2024 / MediQ: Question-Asking LLMs and a Benchmark for Reliable Interactive Clinical Reasoning

MediQ: Question-Asking LLMs and a Benchmark for Reliable Interactive Clinical Reasoning

Paper · 2024 · Shuyue Stella Li, Vidhisha Balachandran, Shangbin Feng et al.

Summary

Most medical QA benchmarks give the model the whole case at once. Real consultations are different: patients often leave out details, and a clinician has to ask follow-up questions before deciding. The authors argue that LLMs are tuned to answer no matter what, which makes them unreliable when the context is incomplete. They introduce MediQ, an interactive benchmark built from MedQA and CRAFT-MD. Each patient record is cut back to a brief intake description (age, gender, chief complaint). A simulated Patient system holds the full record and answers questions. An Expert system must either commit to a multiple-choice answer or ask for more information.

When SOTA models (Llama-3, GPT-3.5, GPT-4) are simply prompted to ask questions, they do worse than when they answer straight away from the limited intake. The proposed MediQ-Expert adds abstention driven by confidence estimation, which recovers much of the lost accuracy. It still falls short of the full-information setting. The work shows that information-seeking is a distinct, unsolved capability that matters for safe use in high-stakes domains.

Key points

  • Conversion pipeline: The pipeline turns single-turn datasets (MedQA, CRAFT-MD) into interactive ones. The Patient system, Expert system, code and data are released on GitHub and Hugging Face (stellalisy/mediQ). The paper appeared at NeurIPS 2024.
  • Patient system variants: Three versions were evaluated for factuality and relevance:
    • Direct, which treats answering as reading comprehension;
    • Instruct, which answers truthfully and declines when the record lacks the answer;
    • Fact-Select, which splits the record into atomic facts and returns the relevant ones.
  • MediQ-Expert design: The system has five modular stages: initial assessment, abstention, question generation, information integration and decision making.
  • Main results:
    • Naively prompting models to ask questions lowered accuracy by 11.3% compared with asking nothing.
    • Abstention strategies such as rationale generation and self-consistency raised accuracy by 22.3%.
    • A 10.3% gap remains relative to the full-information upper bound.
  • Limitations: The benchmark is restricted to multiple-choice questions. Suitable datasets are scarce, and the Patient system depends on a paid API. The authors stress that MediQ is a research prototype and is not intended for deployment.

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

📚 Shelves 5 Computer Science · 9,086 works Artificial Intelligence · 8,889 works Computer Vision and Pattern Recognition · 1,282 works Medicine · 80 works Health Informatics · 53 works

🧭 Topics 2 Reasoning · 510 works Evaluation & benchmarks · 685 works

Cite

Cite as Shuyue Stella Li, Vidhisha Balachandran, Shangbin Feng, Jonathan S. Ilgen, Emma Pierson, Pang Wei Koh and Yulia Tsvetkov. 2024. MediQ: Question-Asking LLMs and a Benchmark for Reliable Interactive Clinical Reasoning. arXiv:2406.00922.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.48550/arxiv.2406.00922 · arXiv 2406.00922

@misc{li2024mediq,
  title = {{MediQ: Question-Asking LLMs and a Benchmark for Reliable Interactive Clinical Reasoning}},
  author = {Shuyue Stella Li, Vidhisha Balachandran, Shangbin Feng, Jonathan S. Ilgen, Emma Pierson, Pang Wei Koh and Yulia Tsvetkov},
  year = {2024},
  doi = {10.48550/arxiv.2406.00922},
  eprint = {2406.00922},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.2406.00922}
}

References

60 references, 23 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Sahar Abdelnabi, Amr Gomaa, Sarath Sivaprasad, Lea Schönherr, and Mario Fritz. Llm-deliberation: Evaluating llms with interactive multi-agent negotiation games, 2023. find
  2. Chinmaya Andukuri, Jan-Philipp Fränken, Tobias Gerstenberg, and Noah D. Goodman. Star-gate: Teaching language models to ask clarifying questions, 2024. find
  3. E. Bolton, D. Hall, M. Yasunaga, T. Lee, C. Manning, and P Liang. Stanford crfm introduces pubmedgpt 2.7b, 2022. URL https://hai.stanford.edu/news/ stanford-crfm-introduces-pubmedgpt-27b. ↗
  4. Georges Bordage. Why did i miss the diagnosis? some cognitive explanations and educational implications. Academic Medicine, 74(10):S138–43, 1999. find
  5. Brian H Bornstein and A Christine Emler. Rationality in medical decision making: a review of the literature on doctors’ decision-making biases. Journal of evaluation in clinical practice, 7(2): 97–107, 2001. find
  6. Henny PA Boshuizen and Henk G Schmidt. On the role of biomedical knowledge in clinical reasoning by experts, intermediates and novices. Cognitive science, 16(2):153–184, 1992. find
  7. Tom B. Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, Sandhini Agarwal, Ariel Herbert-Voss, Gretchen Krueger, Tom Henighan, Rewon Child, Aditya Ramesh, Daniel M. Ziegler, Jeffrey Wu, Clemens Winter,… our card
  8. Chayan Chakrabarti and George F Luger. Artificial conversations for customer service chatter bots: Architecture, algorithms, and evaluation metrics. Expert Systems with Applications, 42(20): 6878–6897, 2015. find
  9. Zeming Chen, Alejandro Hernández Cano, Angelika Romanou, Antoine Bonnet, Kyle Matoba, Francesco Salvi, Matteo Pagliardini, Simin Fan, Andreas Köpf, Amirkeivan Mohtashami, Alexandre Sallinen, Alireza Sakhaeirad, Vinitra Swamy, Igor Krawczuk, Deniz Bayazit, Axel Marmet, Syrielle Montariol, Mary-Anne H… our card
  10. Yang Deng, Lizi Liao, Zhonghua Zheng, Grace Hui Yang, and Tat-Seng Chua. Towards human-centered proactive conversational agents, 2024. find
  11. Arsene Fansi Tchango, Rishab Goel, Zhi Wen, Julien Martel, and Joumana Ghosn. Ddxplus: A new dataset for automatic medical diagnosis. Advances in Neural Information Processing Systems, 35: 31306–31318, 2022. find
  12. Shangbin Feng, Weijia Shi, Yuyang Bai, Vidhisha Balachandran, Tianxing He, and Yulia Tsvetkov. Knowledge card: Filling llms’ knowledge gaps with plug-in specialized language models. In The Twelfth International Conference on Learning Representations, 2023. find
  13. Shangbin Feng, Weijia Shi, Yike Wang, Wenxuan Ding, Vidhisha Balachandran, and Yulia Tsvetkov. Don’t hallucinate, abstain: Identifying llm knowledge gaps via multi-llm collaboration, 2024. find
  14. Jiahui Geng, Fengyu Cai, Yuxia Wang, Heinz Koeppl, Preslav Nakov, and Iryna Gurevych. A survey of language model confidence estimation and calibration. arXiv preprint arXiv:2311.08298, 2023. find
  15. Tianyu Han, Lisa C. Adams, Jens-Michalis Papaioannou, Paul Grundmann, Tom Oberhauser, Alexander Löser, Daniel Truhn, and Keno K. Bressem. Medalpaca – an open-source collection of medical conversational ai models and training data, 2023. our card
  16. He He, Derek Chen, Anusha Balakrishnan, and Percy Liang. Decoupling strategy and generation in negotiation dialogues, 2018. our card
  17. Dan Hendrycks, Collin Burns, Steven Basart, Andy Zou, Mantas Mazeika, Dawn Song, and Jacob Steinhardt. Measuring massive multitask language understanding. arXiv preprint arXiv:2009.03300, 2020. our card
  18. Zhiyuan Hu, Chumin Liu, Xidong Feng, Yilun Zhao, See-Kiong Ng, Anh Tuan Luu, Junxian He, Pang Wei Koh, and Bryan Hooi. Uncertainty of thoughts: Uncertainty-aware planning enhances information seeking in large language models. arXiv preprint arXiv:2402.03271, 2024. find
  19. Wei Huang, Xudong Ma, Haotong Qin, Xingyu Zheng, Chengtao Lv, Hong Chen, Jie Luo, Xiaojuan Qi, Xianglong Liu, and Michele Magno. How good are low-bit quantized llama3 models? an empirical study, 2024. find
  20. Di Jin, Eileen Pan, Nassim Oufattole, Wei-Hung Weng, Hanyi Fang, and Peter Szolovits. What disease does this patient have? a large-scale open domain question answering dataset from medical exams. Applied Sciences, 11(14):6421, 2021. our card
  21. Qiao Jin, Bhuwan Dhingra, Zhengping Liu, William Cohen, and Xinghua Lu. PubMedQA: A dataset for biomedical research question answering. In Kentaro Inui, Jing Jiang, Vincent Ng, and Xiaojun Wan (eds.), Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th… our card
  22. Shreya Johri, Jaehwan Jeong, Benjamin A Tran, Daniel I Schlessinger, Shannon Wongvibulsin, Zhuo Ran Cai, Roxana Daneshjou, and Pranav Rajpurkar. Testing the limits of language models: A conversational framework for medical ai assessmentr. medRxiv, 2023. find
  23. Shreya Johri, Jaehwan Jeong, Benjamin A. Tran, Daniel I Schlessinger, Shannon Wongvibulsin, Zhuo Ran Cai, Roxana Daneshjou, and Pranav Rajpurkar. CRAFT-MD: A conversational evaluation framework for comprehensive assessment of clinical LLMs. In AAAI 2024 Spring Symposium on Clinical Foundation Models… find
  24. Da JU, Song Jiang, Andrew Cohen, Aaron Foss, Sasha Mitts, Arman Zharmagambetov, Brandon Amos, Xian Li, Justine T Kao, Maryam Fazel-Zarandi, and Yuandong Tian. To the globe (ttg): Towards language-driven guaranteed travel planning, 2024. URL https://arxiv.org/abs/ 2410.16456. ↗
  25. Yubin Kim, Chanwoo Park, Hyewon Jeong, Yik Siu Chan, Xuhai Xu, Daniel McDuff, Hyeonhoon Lee, Marzyeh Ghassemi, Cynthia Breazeal, and Hae Won Park. Mdagents: An adaptive collaboration of llms for medical decision-making, 2024. URL https://arxiv.org/abs/2404.15155. — [arXiv:2404.15155](https://arxiv.org/abs/2404.15155) ↗
  26. Belinda Z Li, Alex Tamkin, Noah Goodman, and Jacob Andreas. Eliciting human preferences with language models. arXiv preprint arXiv:2310.11589, 2023. find
  27. Dongdong Li, Zhaochun Ren, Pengjie Ren, Zhumin Chen, Miao Fan, Jun Ma, and Maarten de Rijke. Semi-supervised variational reasoning for medical dialogue generation. In Proceedings of the 44th International ACM SIGIR Conference on Research and Development in Information Retrieval, pp. 544–554, 2021. find
  28. Jessy Lin, Nicholas Tomlin, Jacob Andreas, and Jason Eisner. Decision-oriented dialogue for human-ai collaboration, 2024. find
  29. Ana Marasović, Iz Beltagy, Doug Downey, and Matthew E Peters. Few-shot self-rationalization with natural language prompts. arXiv preprint arXiv:2111.08284, 2021. our card
  30. Izet Masic. Medical decision making-an overview. Acta Informatica Medica, 30(3):230, 2022. find
  31. Sewon Min, Kalpesh Krishna, Xinxi Lyu, Mike Lewis, Wen-tau Yih, Pang Wei Koh, Mohit Iyyer, Luke Zettlemoyer, and Hannaneh Hajishirzi. Factscore: Fine-grained atomic evaluation of factual precision in long form text generation. arXiv preprint arXiv:2305.14251, 2023. our card
  32. Geoffrey Norman. Research in clinical reasoning: past history and current trends. Medical education, 39(4):418–427, 2005. find
  33. 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, Suchir Balaji, Valerie Balcom, Paul Baltescu, Haiming Bao, Mohammad Bavarian, Jeff Belgum, Irwan Bello, Ja… our card
  34. Long Ouyang, Jeff Wu, Xu Jiang, Diogo Almeida, Carroll L. Wainwright, Pamela Mishkin, Chong Zhang, Sandhini Agarwal, Katarina Slama, Alex Ray, John Schulman, Jacob Hilton, Fraser Kelton, Luke Miller, Maddie Simens, Amanda Askell, Peter Welinder, Paul Christiano, Jan Leike, and Ryan Lowe. Training la… our card
  35. Ankit Pal, Logesh Kumar Umapathi, and Malaikannan Sankarasubbu. Medmcqa: A large-scale multi-subject multi-choice dataset for medical domain question answering. In Conference on health, inference, and learning, pp. 248–260. PMLR, 2022. our card
  36. Vimla L Patel, José F Arocha, and David R Kaufman. Diagnostic reasoning and medical expertise. In Psychology of learning and motivation, volume 31, pp. 187–252. Elsevier, 1994. find
  37. Vipula Rawte, Amit Sheth, and Amitava Das. A survey of hallucination in large foundation models. arXiv preprint arXiv:2309.05922, 2023. our card
  38. Henk G Schmidt, Geoffrey R Norman, and Henny P Boshuizen. A cognitive perspective on medical expertise: theory and implication [published erratum appears in acad med 1992 apr; 67 (4): 287]. Academic medicine, 65(10):611–21, 1990. find
  39. Melanie Sclar, Yejin Choi, Yulia Tsvetkov, and Alane Suhr. Quantifying language models’ sensitivity to spurious features in prompt design or: How i learned to start worrying about prompt formatting, 2023. find
  40. Satwinder Singh and Himanshu Beniwal. A survey on near-human conversational agents. Journal of King Saud University-Computer and Information Sciences, 34(10):8852–8866, 2022. find
  41. Karan Singhal, Shekoofeh Azizi, Tao Tu, S Sara Mahdavi, Jason Wei, Hyung Won Chung, Nathan Scales, Ajay Tanwani, Heather Cole-Lewis, Stephen Pfohl, et al. Large language models encode clinical knowledge. Nature, 620(7972):172–180, 2023a. our card
  42. Karan Singhal, Tao Tu, Juraj Gottweis, Rory Sayres, Ellery Wulczyn, Le Hou, Kevin Clark, Stephen Pfohl, Heather Cole-Lewis, Darlene Neal, Mike Schaekermann, Amy Wang, Mohamed Amin, Sami Lachgar, Philip Mansfield, Sushant Prakash, Bradley Green, Ewa Dominowska, Blaise Aguera y Arcas, Nenad Tomasev, Y… our card
  43. Aarohi Srivastava, Abhinav Rastogi, Abhishek Rao, Abu Awal Md Shoeb, Abubakar Abid, Adam Fisch, Adam R Brown, Adam Santoro, Aditya Gupta, Adrià Garriga-Alonso, et al. Beyond the imitation game: Quantifying and extrapolating the capabilities of language models. arXiv preprint arXiv:2206.04615, 2022. our card
  44. Yashar Talebirad and Amirhossein Nadiri. Multi-agent collaboration: Harnessing the power of intelligent llm agents. arXiv preprint arXiv:2306.03314, 2023. find
  45. Katherine Tian, Eric Mitchell, Allan Zhou, Archit Sharma, Rafael Rafailov, Huaxiu Yao, Chelsea Finn, and Christopher D Manning. Just ask for calibration: Strategies for eliciting calibrated confidence scores from language models fine-tuned with human feedback. arXiv preprint arXiv:2305.14975, 2023. find
  46. Hugo Touvron, Louis Martin, Kevin Stone, Peter Albert, Amjad Almahairi, Yasmine Babaei, Nikolay Bashlykov, Soumya Batra, Prajjwal Bhargava, Shruti Bhosale, Dan Bikel, Lukas Blecher, Cristian Canton Ferrer, Moya Chen, Guillem Cucurull, David Esiobu, Jude Fernandes, Jeremy Fu, Wenyin Fu, Brian Fuller,… find
  47. Michael Trimble and Paul Hamilton. The thinking doctor: clinical decision making in contemporary medicine. Clinical Medicine, 16(4):343, 2016. find
  48. Tao Tu, Anil Palepu, Mike Schaekermann, Khaled Saab, Jan Freyberg, Ryutaro Tanno, Amy Wang, Brenna Li, Mohamed Amin, Nenad Tomasev, Shekoofeh Azizi, Karan Singhal, Yong Cheng, Le Hou, Albert Webson, Kavita Kulkarni, S Sara Mahdavi, Christopher Semturs, Juraj Gottweis, Joelle Barral, Katherine Chou,… find
  49. Logesh Kumar Umapathi, Ankit Pal, and Malaikannan Sankarasubbu. Med-halt: Medical domain hallucination test for large language models. arXiv preprint arXiv:2307.15343, 2023. our card
  50. Xuezhi Wang, Jason Wei, Dale Schuurmans, Quoc Le, Ed Chi, Sharan Narang, Aakanksha Chowdhery, and Denny Zhou. Self-consistency improves chain of thought reasoning in language models. arXiv preprint arXiv:2203.11171, 2022. our card
  51. Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Fei Xia, Ed Chi, Quoc V Le, Denny Zhou, et al. Chain-of-thought prompting elicits reasoning in large language models. Advances in neural information processing systems, 35:24824–24837, 2022. our card
  52. Chaoyi Wu, Weixiong Lin, Xiaoman Zhang, Ya Zhang, Yanfeng Wang, and Weidi Xie. Pmc-llama: Towards building open-source language models for medicine. arXiv preprint arXiv:2305.10415, 6, 2023a. our card
  53. Cheng-Kuang Wu, Wei-Lin Chen, and Hsin-Hsi Chen. Large language models perform diagnostic reasoning, 2023b. find
  54. Guande Wu, Chen Zhao, Claudio Silva, and He He. Your co-workers matter: Evaluating collaborative capabilities of language models in blocks world, 2024. find
  55. Qingyun Wu, Gagan Bansal, Jieyu Zhang, Yiran Wu, Shaokun Zhang, Erkang Zhu, Beibin Li, Li Jiang, Xiaoyun Zhang, and Chi Wang. Autogen: Enabling next-gen llm applications via multi-agent conversation framework. arXiv preprint arXiv:2308.08155, 2023c. our card
  56. Johnathan Xie, Annie S Chen, Yoonho Lee, Eric Mitchell, and Chelsea Finn. Calibrating language models with adaptive temperature scaling. In ICLR 2024 Workshop on Secure and Trustworthy Large Language Models, 2024. URL https://openreview.net/forum?id=BgfGqNpoMi. ↗
  57. Ruixin Yang, Dheeraj Rajagopa, Shirley Anugrah Hayati, Bin Hu, and Dongyeop Kang. Confidence calibration and rationalization for llms via multi-agent deliberation. arXiv preprint arXiv:2404.09127, 2024. — [arXiv:2404.09127](https://arxiv.org/abs/2404.09127) find
  58. Michihiro Yasunaga, Antoine Bosselut, Hongyu Ren, Xikun Zhang, Christopher D. Manning, Percy Liang, and Jure Leskovec. Deep bidirectional language-knowledge graph pretraining. ArXiv, abs/2210.09338, 2022. URL https://api.semanticscholar.org/CorpusID:252968266. — [arXiv:2210.09338](https://arxiv.org/abs/2210.09338) ↗
  59. Li Yunxiang, Li Zihan, Zhang Kai, Dan Ruilong, and Zhang You. Chatdoctor: A medical chat model fine-tuned on llama model using medical domain knowledge. arXiv preprint arXiv:2303.14070, 2023. our card
  60. Xuhui Zhou, Hao Zhu, Leena Mathur, Ruohong Zhang, Haofei Yu, Zhengyang Qi, Louis-Philippe Morency, Yonatan Bisk, Daniel Fried, Graham Neubig, and Maarten Sap. Sotopia: Interactive evaluation for social intelligence in language agents, 2024. our card

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/li-2024-mediq-question-asking-llms-benchmark/