2024 AgentHospitalASimulacrumofHospi
- (Li et al., 2024) ⇒ Junkai Li, Siyu Wang, Meng Zhang, Weitao Li, Yunghwei Lai, Xinhui Kang, Weizhi Ma, and Yang Liu. (2024). “Agent Hospital: A Simulacrum of Hospital with Evolvable Medical Agents.” doi:10.48550/arXiv.2405.02957
Subject Headings:
Notes
- The paper introduces Agent Hospital, a simulacrum of a hospital where patients, nurses, and doctors are autonomous agents powered by large language models (LLMs).
- The paper simulates the entire closed cycle of treating a patient's illness, including disease onset, triage, registration, consultation, examination, diagnosis, treatment, convalescence, and follow-up.
- The paper finds that doctor agents can continuously improve their treatment performance over time in the simulation without manually labeled data.
- The paper proposes the MedAgent-Zero strategy, which enables doctor agents to evolve and improve their medical skills by practicing on simulated patients and learning from medical documents.
- The paper demonstrates that the knowledge acquired by doctor agents in Agent Hospital is applicable to real-world medical benchmarks, achieving state-of-the-art accuracy on a subset of the MedQA dataset.
- The paper designs the Agent Hospital environment with various areas like triage, consultation rooms, examination rooms, and includes roles like doctors, nurses, and patient agents.
- The paper simulates patient events and interactions in detail, including disease onset, triage, registration, consultation, examination, diagnosis, treatment recommendation, and convalescence.
- The paper enables doctor agents to practice by treating simulated patients and accumulate experience, as well as learn by reading medical documents in their off-hours.
- The paper evaluates doctor agents on simulated medical tasks like examination decision, diagnosis, and treatment plan recommendation, showing continuous improvement with more simulated patient interactions.
- The paper demonstrates the potential of using comprehensive simulation environments with self-evolving agents to advance AI applications in complex domains like healthcare.
Cited By
Quotes
Abstract
In this paper, we introduce a simulacrum of hospital called Agent Hospital that simulates the entire process of treating illness. All patients, nurses, and doctors are autonomous agents powered by large language models (LLM)s. Our central goal is to enable a doctor agent to learn how to treat illness within the simulacrum. To do so, we propose a method called MedAgent-Zero. As the simulacrum can simulate disease onset and progression based on knowledge bases and LLMs, doctor agents can keep accumulating experience from both successful and unsuccessful cases. Simulation experiments show that the treatment performance of doctor agents consistently improves on various tasks. More interestingly, the knowledge the doctor agents have acquired in Agent Hospital is applicable to real-world medicare benchmarks. After treating around ten thousand patients (real-world doctors may take over two years), the evolved doctor agent achieves a state-of-the-art accuracy of 93.06% on a subset of the MedQA dataset that covers major respiratory diseases. This work paves the way for advancing the applications of LLM-powered agent techniques in medical scenarios.
References
;
| Author | volume | Date Value | title | type | journal | titleUrl | doi | note | year | |
|---|---|---|---|---|---|---|---|---|---|---|
| 2024 AgentHospitalASimulacrumofHospi | Yang Liu Junkai Li Siyu Wang Meng Zhang Weitao Li Yunghwei Lai Xinhui Kang Weizhi Ma | Agent Hospital: A Simulacrum of Hospital with Evolvable Medical Agents | 10.48550/arXiv.2405.02957 | 2024 |