Pith. sign in

REVIEW 3 major objections 5 minor 201 references

ResidencyRL: Reinforcement Learning in Simulated Clinical Environments

T0 review · 3 major / 5 minor · reviewed 2026-08-10 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read Training a medical LLM across full simulated patient encounters lifts adversarial diagnostic accuracy from 81% to 88% and cuts missed red flags by a third.

desk verdict A large, unusually honest RL-for-clinical-dialogue paper whose headline in-domain numbers are reward-optimization checks, but which carries real external evidence led by a strong blinded clinician preference. read the letter →

arxiv 2608.07418 v1 pith:5QRCPXZL submitted 2026-08-07 cs.AI cs.CL

Valentin Liévin , Samuel Schmidgall , Tim Strother , Alex Bijamov , Akshay Goel , Anil Palepu , Chunjong Park , Vahid Balazadeh
show 27 more authors
Min Woo Sun Marius Guerard Justin Chen Dave Steiner Vikram Dhillon Ibrahim Azar Akhil Mehta Nicholas Spetsieris Shilpan Shah Maen Abdelrahim Amit Dahiya Yun Liu Katherine Chou Yossi Matias Avinatan Hassidim Dale R. Webster Quoc V. Le Raia Hadsell Joelle Barral Carey Radebaugh Aleksandra Faust Shekoofeh Azizi Mike Schaekermann Po-Hsuan Cameron Chen Tao Tu David Racz Lin Yang
This is my paper · ORCID
classification cs.AIcs.CL
keywords reinforcementlearningclinicalAIsimulatedpatientencountersmulti-turndialoguediagnosticaccuracyprematureclosureLLMjudgerewardGRPO
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

ResidencyRL claims that the gap between a knowledge-rich medical LLM and a competent clinician is a process gap, and that process can be learned by practicing on simulated patients. The paper trains a base conversational agent through up to 60-turn structured clinical encounters — history-taking, differential diagnosis, management plans, and documentation — against a patient simulator that can hide critical facts or behave adversarially. Training uses online multi-turn reinforcement learning (GRPO) with a reward from an LLM judge that scores diagnosis, management, intake, communication, documentation, and safety. On held-out adversarial cases, the trained agent reaches 88.0% diagnostic accuracy versus 81.0% for the base model, and missed red flags fall from 45.5% to 31.5%; blinded board-certified clinicians prefer it in 87.6% of side-by-side comparisons. The paper's central claim is that sequential clinical decision-making can be effectively learned in simulation, yielding robust, generalizable procedural competencies.

What carries the argument

The load-bearing mechanism is online group relative policy optimization (GRPO), an RL update that estimates advantages from a group of parallel rollouts without a value network, executed over a simulated clinical POMDP with a trajectory-level structured reward $R = R_{\text{primary}} - R_{\text{penalty}} \in [-3, 3]$. The reward is produced by an LLM judge that decomposes the encounter into six weighted clinical dimensions (diagnosis $2/9$, management $3/9$, intake $1/9$, communication $1/9$, documentation $1/9$, style $1/9$) spanning 26 Likert sub-axes and up to 8 binary safety flags, with penalties for hallucinations, contraindicated actions, missed critical questions, and under-triage. The patient simulator, conditioned on curated scenario context, withholds critical facts until the agent asks specifically and can sustain deception across turns, so the policy must actively probe rather than accept surface presentation. A scenario generation pipeline (roughly 57K cases) grounds the curriculum in curated evidence and layers targeted history-taking and adversarial safety scenarios.

What would settle it

Run the trained and base agents with the reward and evaluation assigned by a different judge — for example, a separately calibrated human-preference model or a different LLM family — and compare in-domain diagnostic accuracy and missed-red-flag rates, while confirming the base model's scores on the original rubric are unchanged. If the trained agent's advantage shrinks or disappears, the gains are at least partly artifacts of the specific judge rather than genuine clinical skill; a complementary check is deploying the trained agent in a human standardized-patient setting and measuring whether the completeness-of-information-gathering advantage still appears.

Watch

Extended reading notes

Core claim

In the authors' own terms, the discovery is that a frontier model that already possesses broad medical knowledge can be made into a meaningfully more competent and safer clinician through multi-turn reinforcement learning in simulation. The trained agent learns when to keep gathering information, when to escalate care, how to document without fabricating, and how to manage clinical uncertainty — skills that single-turn benchmarks and per-turn supervision do not teach. The evidence is the consistent pattern of improvement across adversarial safety cases, longitudinal multi-visit care, specialist oncology consultations, and external benchmarks, all favoring the trained agent over its base model.

Load-bearing premise

The entire result rests on the assumption that the LLM judge used as both the training reward and the in-domain evaluation metric measures genuine clinical quality without being gamed, even though the paper documents a systematic positive bias of that judge toward the trained model.

Editorial extensions

If this is right

  • If the central claim is right, long-horizon multi-turn RL becomes an available training method for clinical dialogue, complementing single-turn medical benchmarks with process-level optimization.
  • Adversarial safety training in simulation translates to measurable safety gains: the trained agent misses roughly one-third fewer red flags and critical questions on held-out adversarial encounters.
  • The learned procedural competencies transfer out of domain: all six clinical axes improve on the AMIE Mx multi-visit benchmark, and oncology, AgentClinic, and CRAFT-MD show directional or significant gains.
  • The trained agent remains better than the base model even inside an expert-optimized agentic harness, meaning simulation-based RL adds value on top of careful engineering.
  • Blinded clinicians prefer the trained agent in 87.6% of comparisons, which would imply that the automated judge metrics track clinically meaningful quality rather than only rubric artifacts.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • A testable extension the paper leaves open: swapping the reward and evaluation judge for an independently calibrated one (a different model family or a human-preference signal) would separate genuine process learning from exploitation of judge artifacts, since the paper documents a systematic positive bias of the judge toward the trained model.
  • The training curriculum expands naturally toward multi-visit and multi-actor encounters; if the pattern holds, training agents to observe the downstream effects of their own management decisions should further improve longitudinal care, a prediction that could be tested by extending the horizon beyond single visits.
  • The framework suggests an empirical scaling relation for clinical simulation: encounter horizon, scenario diversity, and adversarial difficulty jointly set the competence ceiling, so reporting gains at longer horizons would corroborate that the mechanism, not the specific rubric, is what drives improvement.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper presents ResidencyRL, an online GRPO training method that optimizes a Gemini 3.5 Flash–based agent over complete simulated clinical encounters, using Gemini 3.1 Pro as the patient simulator, scenario generator, and structured autorater. The authors report in-domain gains in diagnostic accuracy, management quality, communication, and safety, plus transfer to AMIE Mx, expert oncology cases, AgentClinic, CRAFT-MD, and an out-of-domain telehealth harness; a blinded clinician study prefers the trained agent in 87.6% of cases. The abstract's flagship 7.0 pp diagnostic-accuracy gain is measured by the same autorater that supplied the training reward, a circularity the paper acknowledges in Section 4.1 and Appendix C.7.

Significance. The paper makes a credible case that multi-turn RL with a rich action space (dialogue plus a documentation API, up to 68 actions) and adversarial simulated patients can improve an LLM's clinical process, not only its knowledge. Strengths include an explicit and detailed reward decomposition (Table 2), a scenario-generation pipeline with verification and deduplication, an in-context-learning ablation (Appendix E) showing RL outperforms demonstration-only exposure, and a blinded clinician side-by-side evaluation (Section 4.5) with a strong preference for the trained agent. The manuscript also deserves credit for transparently acknowledging the reward-judge circularity and the autorater's positive bias (Section 4.1, Discussion, Appendix C.7), which makes the remaining issues fixable. If the reported effects hold under independent evaluation, this would be a meaningful advance in agentic clinical RL.

major comments (3)
  1. [§4.1; §3.3–3.4; Appendix C.7] The headline diagnostic-accuracy improvement (81.0% → 88.0% on adversarial cases, abstract and Section 4.1) is produced by the same Gemini 3.1 Pro rubric pipeline that acts as the GRPO training reward (Sections 3.3–3.4). Section 4.1 itself states that 'all evaluations used the same automated rubric pipeline that served as the training reward signal' and that these results 'primarily validate that the agent successfully learns to optimize its training signals.' Appendix C.7 additionally documents a systematic positive autorater delta toward the trained model (+0.63 overall, including cases where clinicians did not prefer it). Since the abstract and introduction present this in-domain number as evidence of improved diagnostic accuracy without that caveat, the manuscript's central quantitative claim is vulnerable to judge exploitation. Please either report in-domain numbers explicitly as reward-optimization checks, or re-evaluate a held-out subset with an independent judge and/or human diagnostic grading.
  2. [§4.4, Table 6] The generalization claim rests on non-significant differences: AgentClinic-MedQA p=0.176, AgentClinic-MIMIC-IV p=0.079, and all CRAFT-MD comparisons have overlapping confidence intervals. The text argues from 'consistent directional improvements,' but with eight comparisons this pattern is expected by chance under the null. Provide a pre-specified primary analysis, a combined test across benchmarks (e.g., Fisher's method), or multiplicity adjustment; without this, the 'procedural competencies transfer' conclusion in the abstract and Section 1 is not statistically supported. The AMIE Mx and human side-by-side results are the stronger transfer evidence, but those involve Gemini-based rubric scoring (Section 4.2) or clinician preference rather than independent diagnostic labels (Section 4.5).
  3. [§4.2, §4.5, Appendix C.7] AMIE Mx is scored by a Gemini 3.1 Pro autorater, the same model that provides the training reward. Given the documented bias pattern in Appendix C.7, it would be valuable to show that the large AMIE Mx gains (e.g., +8.34 pp on MXEKF) are not partly judge-induced; for example, re-score a random subset of AMIE Mx transcripts with the judge blinded to model identity, or calibrate the autorater deltas against clinician ratings on that benchmark. Relatedly, the human side-by-side 'Diagnostic Assessment' win rate (66.0%, Section 4.5) is considerably lower than the abstract's 7.0 pp diagnostic-accuracy gain, and the manuscript does not reconcile these two constructs. Reporting the human diagnostic assessment as the primary diagnostic outcome would make the claim more robust.
minor comments (5)
  1. [Abstract; §4.1] The abstract's 'reduces missed red flag rates by 31%' is a relative reduction (14.0/45.5 = 30.8%); state whether the value is relative or absolute to avoid misinterpretation.
  2. [Appendix C.7] The phrase 'webinarized the comparison' should read 'binarized the comparison'.
  3. [§4.5] The text reports '99/100 encounter cases are judged as realistic' but 3 of 100 sampled scenarios were excluded, leaving n=97; specify the denominator and whether realism ratings refer to the full 100 or the analyzed 97.
  4. [§4.3] The qualitative oncologist review of N=100 cases reports no number of reviewing oncologists, no inter-rater agreement, and no formal analysis; adding these details would allow readers to gauge reliability.
  5. [Appendix B, Table A5] Table A5 gives anchors for scores 5, 3, and 1 on the Diagnosis dimension but no anchor for 4, while Section 4.1 defines ≥4/5 as 'Good differential including correct diagnosis with minor omissions'; please add the score-4 anchor for consistency.

Circularity Check

1 steps flagged · score 4.0 of 10

In-domain diagnostic and safety gains are measured by the same Gemini 3.1 Pro autorater that served as the GRPO reward; the paper's own caveat and Appendix C.7 confirm a positive bias toward the trained model, though independent clinician side-by-side evaluation partially grounds the central claim.

  1. fitted input called prediction [Section 4.1 / Table 4; Section 3.3-3.4; Appendix C.7]
    "All evaluations used the same automated rubric pipeline that served as the training reward signal. ... At each episode completion, Gemini 3.1 Pro evaluates the conversation transcript and documentation submitted via tool calls against the ground-truth diagnosis and management, yielding a trajectory-level scalar reward. ..."

    The headline diagnostic-accuracy gain (81.0% to 88.0% on adversarial cases) and the missed-red-flag reduction (45.5% to 31.5%) are computed by the same Gemini 3.1 Pro rubric pipeline that served as the training reward in GRPO (Sections 3.3 and 3.4). The policy was explicitly optimized to maximize that judge's scores, so reporting held-out improvement on the same judge is, by construction, reporting improvement on the training objective rather than an independent measure of clinical accuracy.

full rationale

The paper is not globally circular: multi-turn RL against an LLM-judge reward is a legitimate method, and the authors are unusually explicit that the in-domain numbers primarily validate optimization of the training signal. The specific circularity is confined to presenting those in-domain rubric gains as clinical diagnostic-accuracy improvements without the caveat, since the Gemini 3.1 Pro autorater was both the training reward and the evaluation metric. Appendix C.7 confirms a systematic positive bias toward the trained model. However, the central claim retains independent content: blinded clinician side-by-side evaluation (87.6% preference, Section 4.5), AMIE Mx gains on a separately validated rubric, expert-curated oncology cases, and directional external benchmarks on AgentClinic and CRAFT-MD provide evidence not reducible to the reward judge. No load-bearing self-citation chain or uniqueness theorem is present; the AMIE Mx citation is backed by prior physician validation. The external benchmark improvements are individually non-significant, which is a statistical weakness rather than a circularity. Overall, partial circularity in the headline in-domain metric, with meaningful independent grounding elsewhere, supports a score of 4.

Assumptions & free parameters 5 free parameters · 5 assumptions · 0 invented entities

The central claim rests on hand-set reward weights and penalties (Table 2, Eq. 1), a scenario mix (83/15/2), and thresholds (TF-IDF 0.9 and 0.45), plus the validity of LLM-based patient simulation and LLM-as-judge scoring. The paper discloses some limitations but does not ablate these design choices.

free parameters (5)
  • Reward dimension weights = 2/9 diagnosis, 3/9 management, 1/9 each for communication, documentation, intake, style
    Hand-chosen in Eq. 1 to prioritize management and diagnostic accuracy; not empirically tuned or justified.
  • Safety flag penalties = 2.0 hallucination, 3.0 contraindication, 2.0 plan leakage, 1.0 other; additional 2.0/3.0/3.0/3.0 for adversarial flags
    Hand-set magnitudes in Table 2 determine reward optimization pressure; no ablation justifies these values.
  • Turn-length penalty schedule = 0.1 at 30 turns ramping to 1.0 at 40 turns
    Hand-chosen in Section 3.3 to discourage exhaustive symptom enumeration; no sensitivity analysis.
  • Scenario mix proportions = 83% general, 15% history-taking, 2% adversarial
    Chosen in Section 3.4; affects the balance of skills learned by the agent.
  • TF-IDF similarity thresholds = 0.9 deduplication, 0.45 evaluation filtering
    Hand-set thresholds in Section 3.1 and 4.1 control dataset diversity and eval overlap; impact on results not analyzed.
assumptions (5)
  • domain assumption DDXPlus evidence corpus provides clinically valid associations for grounded scenario generation
    Section 3.1 Stage 2 grounds generated scenarios in DDXPlus; if this corpus is biased or incomplete, scenario quality and thus training upper-bound is affected.
  • domain assumption LLM-as-judge (Gemini 3.1 Pro) scores clinical quality on 1-5 Likert scales in a valid, consistent manner
    Used for both reward (Section 3.3) and evaluation (Section 4); the paper acknowledges positive bias (Appendix C.7).
  • domain assumption Big Five personality traits sampled from population-calibrated Gaussians (Stackhouse et al., 2024) validly modulate patient behavior
    Section 3.1 Stage 1; if trait-behavior links are inaccurate, simulated patient diversity and realism is compromised.
  • domain assumption US demographic distributions (CDC NHIS 2021, Census) are appropriate for training a generalizable clinical agent
    Section 3.1 Stage 1; limits to English-speaking US-based patients, as acknowledged in Limitations.
  • standard math GRPO advantage estimation and Wilcoxon/bootstrap statistics are correctly applied
    Standard methods (Shao et al., 2024); no derivations in this paper require novel mathematics.

how reviews work

0 comments
Cite this review

Pith. "Pith review of ResidencyRL: Reinforcement Learning in Simulated Clinical Environments." pith.science (2026). https://pith.science/paper/5QRCPXZL

@misc{pith2026260807418,
  author       = {Pith},
  title        = {Pith review of: ResidencyRL: Reinforcement Learning in Simulated Clinical Environments},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/5QRCPXZL}},
  note         = {Machine review of arXiv:2608.07418}
}
read the original abstract

In medical education, physicians convert academic knowledge into clinical expertise through residency: years of training across thousands of encounters, with diverse sources of feedback and progressively greater autonomy. Much of clinical reasoning relies on the patient encounter, a dialogue in which a clinician elicits history, refines diagnostic hypotheses, and decides management under uncertainty. While large language models (LLMs) excel on static medical benchmarks, methods to optimize the full sequence of clinical decisions remain underdeveloped. We present ResidencyRL, a reinforcement learning (RL) method for training clinical artificial intelligence (AI) agents through simulated multi-turn clinical encounters (up to 60 dialogue turns and 8 tool calls per trajectory). ResidencyRL pairs the policy agent with LLM simulators capable of complex, adversarial behaviors, training against a structured reward aligned to diagnostic accuracy, management quality, communication, documentation, and safety. On held-out evaluations, the ResidencyRL agent improves diagnostic accuracy by 7.0% under adversarial conditions (88.0% vs. 81.0%) and reduces missed red flag rates by 31%, demonstrating rigorous mitigation of premature closure. Blinded expert clinicians validated these gains, preferring the trained agent in 87.6% of side-by-side comparisons. The procedural competencies transfer to unseen benchmarks: the agent outperforms the base model across all six clinical axes of the AMIE multi-visit benchmark, and shows consistent directional improvements on AgentClinic and CRAFT-MD. Our findings demonstrate that sequential clinical decision-making can be effectively learned through multi-turn RL in simulation, yielding robust, generalizable capabilities, paving the way towards clinical mastery. Prospective validation with real-world workflows remains necessary to establish clinical utility.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

201 extracted references · 37 canonical work pages

  1. [1]

    2017 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , pages=

    Domain randomization for transferring deep neural networks from simulation to the real world , author=. 2017 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) , pages=. 2017 , organization=

    2017

  2. [2]

    2018 IEEE International Conference on Robotics and Automation (ICRA) , pages=

    Sim-to-Real Transfer of Robotic Control with Dynamics Randomization , author=. 2018 IEEE International Conference on Robotics and Automation (ICRA) , pages=. 2018 , organization=

    2018

  3. [3]

    arXiv preprint arXiv:2009.13303 , year=

    Sim-to-Real Transfer in Deep Reinforcement Learning for Robotics: a Survey , author=. arXiv preprint arXiv:2009.13303 , year=

    arXiv 2009

  4. [4]

    arXiv preprint arXiv:1910.07113 , year=

    Solving Rubik's Cube with a Robot Hand , author=. arXiv preprint arXiv:1910.07113 , year=

    arXiv 1910

  5. [5]

    Nature , volume=

    A graph placement methodology for fast chip design , author=. Nature , volume=. 2021 , publisher=

    2021

  6. [6]

    and Veness, Joel and Bellemare, Marc G

    Mnih, Volodymyr and Kavukcuoglu, Koray and Silver, David and Rusu, Andrei A. and Veness, Joel and Bellemare, Marc G. and Graves, Alex and Riedmiller, Martin and Fidjeland, Andreas K. and Ostrovski, Georg and Petersen, Stig and Beattie, Charles and Sadik, Amir and Antonoglou, Ioannis and King, Helen and Kumaran, Dharshan and Wierstra, Daan and Legg, Shane ...

  7. [7]

    Learning Navigation Behaviors End-to-End With AutoRL , year=

    Chiang, Hao-Tien Lewis and Faust, Aleksandra and Fiser, Marek and Francis, Anthony , journal=. Learning Navigation Behaviors End-to-End With AutoRL , year=

  8. [8]

    Nature , volume=

    Autonomous navigation of stratospheric balloons using reinforcement learning , author=. Nature , volume=. 2020 , publisher=

    2020

Show all 201 references
  1. [9]

    Environment Generation for Zero-Shot Compositional Reinforcement Learning , url =

    Gur, Izzeddin and Jaques, Natasha and Miao, Yingjie and Choi, Jongwook and Tiwari, Manoj and Lee, Honglak and Faust, Aleksandra , booktitle =. Environment Generation for Zero-Shot Compositional Reinforcement Learning , url =

  2. [10]

    Nature , volume=

    Towards conversational diagnostic artificial intelligence , author=. Nature , volume=. 2025 , publisher=

    2025

  3. [11]

    Advancing conversational diagnostic AI with multimodal reasoning

    Saab, Khaled and Park, Chunjong and Strother, Tim and Freyberg, Jan and Barrett, David G T and Cheng, Yong and Weng, Wei-Hung and Stutz, David and Tomasev, Nenad and Palepu, Anil and Liévin, Valentin and Sharma, Yash and Ruparel, Roma and Ahmed, Abdullah and Vedadi, Elahe and ...

  4. [12]

    Towards physician-centered oversight of conversational diagnostic

    Vedadi, Elahe and Barrett, David and Harris, Natalie and Wulczyn, Ellery and Reddy, Shashir and Ruparel, Roma and Schaekermann, Mike and Strother, Tim and Tanno, Ryutaro and Sharma, Yash and others , journal=. Towards physician-centered oversight of conversational diagnostic

  5. [13]

    Mu, Linjie and Huang, Zhongzhen and Gu, Yannian and Qin, Shengqian and Zhang, Shaoting and Zhang, Xiaofan , journal=

  6. [14]

    2025 , publisher=

    Jiang, Yixing and Black, Kameron C and Geng, Gloria and Park, Danny and Zou, James and Ng, Andrew Y and Chen, Jonathan H , journal=. 2025 , publisher=

    2025

  7. [15]

    The landscape of agentic reinforcement learning for LLMs : A survey

    Zhang, Guibin and Geng, Hejia and Yu, Xiaohang and Yin, Zhenfei and Zhang, Zaibin and Tan, Zelin and Zhou, Heng and Li, Zhongzhi and Xue, Xiangyuan and Li, Yijiang and Zhou, Yifan and Chen, Yang and Zhang, Chen and Fan, Yutao and Wang, Zihu and Huang, Songtao and Piedrahita-Ve...

    doi arXiv

  8. [16]

    Towards conversational AI for disease management

    Liévin, Valentin and Palepu, Anil and Weng, Wei-Hung and Saab, Khaled and Stutz, David and Cheng, Yong and Kulkarni, Kavita and Mahdavi, S Sara and Barral, Joëlle and Webster, Dale R and Chou, Katherine and Hassidim, Avinatan and Matias, Yossi and Manyika, James and Tanno, Ryu...

    doi

  9. [17]

    Nature , volume=

    Towards Accurate Differential Diagnosis with Large Language Models , author=. Nature , volume=. 2025 , publisher=

    2025

  10. [18]

    Nature Medicine , volume=

    A large language model for complex cardiology care , author=. Nature Medicine , volume=. 2026 , publisher=

    2026

  11. [19]

    A prospective clinical feasibility study of a conversational diagnostic

    Brodeur, Peter and Koshy, Jacob M and Palepu, Anil and Saab, Khaled and Homiar, Ava and Ruparel, Roma and Wu, Charles and Tanno, Ryutaro and Xu, Joseph and Wang, Amy and others , journal=. A prospective clinical feasibility study of a conversational diagnostic

  12. [20]

    Can large language models reason about medical questions?

    Liévin, Valentin and Hother, Christoffer Egeberg and Motzfeldt, Andreas Geert and Winther, Ole. Can large language models reason about medical questions?. Patterns (N. Y.). doi:10.1016/j.patter.2024.100943

    2024

  13. [21]

    Nature Medicine , volume=

    Towards Expert-Level Medical Question Answering with Large Language Models , author=. Nature Medicine , volume=. 2025 , publisher=

    2025

  14. [22]

    Capabilities of

    Saab, Khaled and Tu, Tao and Amatriain, Xavier and others , journal=. Capabilities of

  15. [23]

    Schmidgall, Samuel and Ziaei, Rojin and Harris, Carl and others , journal=

  16. [24]

    Evaluation of

    Johri, Sanjat and others , journal=. Evaluation of

  17. [25]

    Jin, Zhexin and others , journal=

  18. [26]

    Sun, Lichao and others , journal=

  19. [27]

    Capabilities of

    Nori, Harsha and King, Nicholas and McKinney, Scott Mayer and Carignan, Dean and Horvitz, Eric , journal=. Capabilities of

  20. [28]

    Mastering the Game of

    Silver, David and Huang, Aja and Maddison, Chris J and others , journal=. Mastering the Game of. 2016 , publisher=

    2016

  21. [29]

    Mastering the Game of

    Silver, David and Schrittwieser, Julian and Simonyan, Karen and others , journal=. Mastering the Game of. 2017 , publisher=

    2017

  22. [30]

    Mastering

    Schrittwieser, Julian and Antonoglou, Ioannis and Hubert, Thomas and others , journal=. Mastering. 2020 , publisher=

    2020

  23. [31]

    Advances in Neural Information Processing Systems , volume=

    Training Language Models to Follow Instructions with Human Feedback , author=. Advances in Neural Information Processing Systems , volume=

  24. [32]

    Xi, Zhiheng and Huang, Jixuan and Liao, Chenyang and Huang, Baodai and Guo, Honglin and Liu, Jiaqi and Zheng, Rui and Ye, Junjie and Zhang, Jiazheng and Chen, Wenxiang and others , journal=

  25. [33]

    Grandmaster level in

    Vinyals, Oriol and Babuschkin, Igor and Czarnecki, Wojciech M and Mathieu, Micha. Grandmaster level in. Nature , volume=. 2019 , publisher=

    2019

  26. [34]

    Nature , volume=

    Discovering faster matrix multiplication algorithms with reinforcement learning , author=. Nature , volume=. 2022 , publisher=

    2022

  27. [35]

    Proceedings of the 33rd Conference on Uncertainty in Artificial Intelligence , year=

    A reinforcement learning approach to weaning of mechanical ventilation in intensive care units , author=. Proceedings of the 33rd Conference on Uncertainty in Artificial Intelligence , year=

  28. [36]

    38th Annual International Conference of the IEEE Engineering in Medicine and Biology Society , pages=

    Optimal medication dosing from suboptimal clinical examples: A deep reinforcement learning approach , author=. 38th Annual International Conference of the IEEE Engineering in Medicine and Biology Society , pages=. 2016 , publisher=

    2016

  29. [37]

    Proceedings of the 2nd Machine Learning for Healthcare Conference , series=

    Continuous state-space models for optimal sepsis treatment: a deep reinforcement learning approach , author=. Proceedings of the 2nd Machine Learning for Healthcare Conference , series=. 2017 , publisher=

    2017

  30. [38]

    Journal of the Royal Statistical Society Series B: Statistical Methodology , volume=

    Optimal Dynamic Treatment Regimes , author=. Journal of the Royal Statistical Society Series B: Statistical Methodology , volume=. 2003 , doi=

    2003

  31. [39]

    Proceedings of the Second Seattle Symposium in Biostatistics , series=

    Optimal Structural Nested Models for Optimal Sequential Decisions , author=. Proceedings of the Second Seattle Symposium in Biostatistics , series=. 2004 , doi=

    2004

  32. [40]

    Annual Review of Statistics and Its Application , volume=

    Dynamic Treatment Regimes , author=. Annual Review of Statistics and Its Application , volume=. 2014 , doi=

    2014

  33. [41]

    Machine Learning , volume=

    Informing Sequential Clinical Decision-Making through Reinforcement Learning: An Empirical Study , author=. Machine Learning , volume=. 2011 , doi=

    2011

  34. [42]

    Clinical Data Based Optimal

    Ernst, Damien and Stan, Guy-Bart and Goncalves, Jorge and Wehenkel, Louis , booktitle=. Clinical Data Based Optimal. 2006 , doi=

    2006

  35. [43]

    Medical Physics , volume=

    Deep Reinforcement Learning for Automated Radiation Adaptation in Lung Cancer , author=. Medical Physics , volume=. 2017 , doi=

    2017

  36. [44]

    Machine Learning for Healthcare Conference , year=

    Reinforcement Learning with Action-Derived Rewards for Chemotherapy and Clinical Trial Dosing Regimen Selection , author=. Machine Learning for Healthcare Conference , year=

  37. [45]

    IEEE Journal of Biomedical and Health Informatics , volume=

    Basal Glucose Control in Type 1 Diabetes Using Deep Reinforcement Learning: An In Silico Validation , author=. IEEE Journal of Biomedical and Health Informatics , volume=. 2020 , doi=

    2020

  38. [46]

    Journal of Medical Internet Research , volume=

    Reinforcement Learning for Clinical Decision Support in Critical Care: Comprehensive Review , author=. Journal of Medical Internet Research , volume=. 2020 , doi=

    2020

  39. [47]

    Artificial Intelligence in Medicine , volume=

    Reinforcement Learning for Intelligent Healthcare Applications: A Survey , author=. Artificial Intelligence in Medicine , volume=. 2020 , doi=

    2020

  40. [48]

    Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing , year=

    Deep Reinforcement Learning for Dialogue Generation , author=. Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing , year=

    2016

  41. [49]

    Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing , year=

    Composite Task-Completion Dialogue Policy Learning via Hierarchical Deep Reinforcement Learning , author=. Proceedings of the 2017 Conference on Empirical Methods in Natural Language Processing , year=

    2017

  42. [50]

    Lai, Yunghwei and Liu, Kaiming and Wang, Ziyue and Ma, Weizhi and Liu, Yang , journal=

  43. [51]

    Academic Medicine , volume=

    Computerized virtual patients in health professions education: a systematic review and meta-analysis , author=. Academic Medicine , volume=. 2010 , publisher=

    2010

  44. [52]

    Personality and Individual Differences , volume=

    Homogeneity, heterogeneity, or independence? A multilevel exploration of Big Five personality traits and cultural values in 40 nations , author=. Personality and Individual Differences , volume=. 2024 , publisher=

    2024

  45. [53]

    Journal of documentation , volume=

    A statistical interpretation of term specificity and its application in retrieval , author=. Journal of documentation , volume=. 1972 , publisher=

    1972

  46. [54]

    Journal of personality , volume=

    Neuroticism, somatic complaints, and disease: is the bark worse than the bite? , author=. Journal of personality , volume=. 1987 , publisher=

    1987

  47. [55]

    Scientific data , volume=

    Demographic aspects of first names , author=. Scientific data , volume=. 2018 , publisher=

    2018

  48. [56]

    Intelligent Virtual Agents: 7th International Conference, IVA 2007, Paris, France, September 17-19, 2007

    Virtual patients for clinical therapist skills training , author=. Intelligent Virtual Agents: 7th International Conference, IVA 2007, Paris, France, September 17-19, 2007. Proceedings 7 , pages=. 2007 , organization=

    2007

  49. [57]

    Nature , volume=

    Large language models encode clinical knowledge , author=. Nature , volume=. 2023 , publisher=

    2023

  50. [58]

    Nejm Ai , volume=

    Towards generalist biomedical AI , author=. Nejm Ai , volume=. 2024 , publisher=

    2024

  51. [59]

    Proceedings of Machine Learning Research (ML4H) , year=

    Med-Flamingo: a Multimodal Medical Few-shot Learner , author=. Proceedings of Machine Learning Research (ML4H) , year=

  52. [60]

    2024 , publisher=

    Almanac—retrieval-augmented language models for clinical medicine , author=. 2024 , publisher=

    2024

  53. [61]

    npj Digital Medicine , volume=

    The imperative for regulatory oversight of large language models (or generative AI) in healthcare , author=. npj Digital Medicine , volume=. 2023 , publisher=

    2023

  54. [62]

    JAMA , volume=

    Accuracy of a generative artificial intelligence model in a complex diagnostic challenge , author=. JAMA , volume=. 2023 , publisher=

    2023

  55. [63]

    arXiv preprint arXiv:2505.00467 , year=

    Red teaming large language models for healthcare , author=. arXiv preprint arXiv:2505.00467 , year=

    arXiv

  56. [64]

    JAMA Network Open , volume=

    Prevalence of and factors associated with patient nondisclosure of medically relevant information to clinicians , author=. JAMA Network Open , volume=. 2018 , publisher=

    2018

  57. [65]

    Scandinavian Journal of Trauma, Resuscitation and Emergency Medicine , volume=

    Nonspecific complaints in the emergency department--a systematic review , author=. Scandinavian Journal of Trauma, Resuscitation and Emergency Medicine , volume=. 2020 , publisher=

    2020

  58. [66]

    International Conference on Machine Learning , pages=

    Variational open-domain question answering , author=. International Conference on Machine Learning , pages=. 2023 , organization=

    2023

  59. [67]

    Conference on Health, Inference, and Learning , pages=

    MedMCQA: A large-scale multi-subject multi-choice dataset for medical domain question answering , author=. Conference on Health, Inference, and Learning , pages=. 2022 , organization=

    2022

  60. [68]

    Medical education online , volume=

    The Dreyfus model of clinical problem-solving skills acquisition: a critical perspective , author=. Medical education online , volume=. 2010 , publisher=

    2010

  61. [69]

    2024 , publisher=

    Wu, Chaoyi and Lin, Weixiong and Zhang, Xiaoman and Zhang, Ya and Xie, Weidi and Wang, Yanfeng , journal=. 2024 , publisher=

    2024

  62. [70]

    Nature medicine , volume=

    Multimodal biomedical AI , author=. Nature medicine , volume=. 2022 , publisher=

    2022

  63. [71]

    NPJ digital medicine , volume=

    Deep learning-enabled medical computer vision , author=. NPJ digital medicine , volume=. 2021 , publisher=

    2021

  64. [72]

    2026 , organization=

    Feng, Yichun and Wang, Jiawei and Zhou, Lu and Lei, Zhen and Li, Yixue , booktitle=. 2026 , organization=

    2026

  65. [73]

    and Fan, Xiaoran and Shangguan, Longfei and Wang, Anran and Daryani, Bhavna and Herkenham, Buddy and Tan, Cara and Malhotra, Mark and Patel, Shwetak and Hernandez, John B

    Breda, Joseph and Yousif, Fadi and Hawkins, Beszel and Cotoi, Marinela and Liu, Miao and Luo, Ray and Chen, Po-Hsuan Cameron and Schaekermann, Mike and Schmidgall, Samuel and Liu, Xin and Narayanswamy, Girish and Solomon, Samuel and Xu, Maxwell A. and Fan, Xiaoran and Shanggua...

  66. [74]

    Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics , year=

    Task-oriented Dialogue System for Automatic Diagnosis , author=. Proceedings of the 56th Annual Meeting of the Association for Computational Linguistics , year=

  67. [75]

    2021 , publisher=

    Development and validation of a reinforcement learning algorithm to dynamically optimize mechanical ventilation in critical care , author=. 2021 , publisher=

    2021

  68. [76]

    NPJ digital medicine , volume=

    A primer on reinforcement learning in medicine for clinicians , author=. NPJ digital medicine , volume=. 2024 , publisher=

    2024

  69. [77]

    Wang, Guangyu and Yang, Guoxing and Du, Zongxin and Fan, Longjun and Li, Xiaohu , journal=

  70. [78]

    arXiv preprint arXiv:2305.12031 , year=

    Clinical camel: An open expert-level medical language model with dialogue-based knowledge encoding , author=. arXiv preprint arXiv:2305.12031 , year=

    arXiv

  71. [79]

    arXiv preprint arXiv:2304.08247 , year=

    Han, Tianyu and Adams, Lisa C and Papaioannou, Jens-Michalis and Grundmann, Paul and Oberhauser, Tom and Figueroa, Alexei and L. arXiv preprint arXiv:2304.08247 , year=

    arXiv

  72. [80]

    NPJ digital medicine , volume=

    Clinical entity augmented retrieval for clinical information extraction , author=. NPJ digital medicine , volume=. 2025 , publisher=

    2025

  73. [81]

    Zhao, Xuejiao and Liu, Siyan and Yang, Su-Yin and Miao, Chunyan , booktitle=

  74. [82]

    Nature , volume=

    Faster sorting algorithms discovered using deep reinforcement learning , author=. Nature , volume=. 2023 , publisher=

    2023

  75. [83]

    2025 , organization=

    Pan, Jiazhen and Liu, Che and Wu, Junde and Liu, Fenglin and Zhu, Jiayuan and Li, Hongwei Bran and Chen, Chen and Ouyang, Cheng and Rueckert, Daniel , booktitle=. 2025 , organization=

    2025

  76. [84]

    arXiv preprint arXiv:2311.16079 , year=

    Meditron-70b: Scaling medical pretraining for large language models , author=. arXiv preprint arXiv:2311.16079 , year=

    arXiv

  77. [85]

    2026 , publisher=

    Lai, Yuxiang and Zhong, Jike and Li, Ming and Zhao, Shitian and Li, Yuheng and Psounis, Konstantinos and Yang, Xiaofeng , journal=. 2026 , publisher=

    2026

  78. [86]

    arXiv preprint arXiv:2604.05081 , year=

    Medgemma 1.5 technical report , author=. arXiv preprint arXiv:2604.05081 , year=

    arXiv

  79. [87]

    arXiv preprint arXiv:2507.05201 , year=

    Medgemma technical report , author=. arXiv preprint arXiv:2507.05201 , year=

    arXiv

  80. [88]

    arXiv preprint arXiv:2204.05862 , year=

    Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback , author=. arXiv preprint arXiv:2204.05862 , year=

    arXiv

  81. [89]

    Constitutional

    Bai, Yuntao and Kadavath, Saurav and Kundu, Sandipan and others , journal=. Constitutional

  82. [90]

    Advances in Neural Information Processing Systems , volume=

    Deep Reinforcement Learning from Human Preferences , author=. Advances in Neural Information Processing Systems , volume=

  83. [91]

    Ramsay, J and Campbell, JL and Schroter, S and Green, J and Roland, M , journal=. The

  84. [92]

    Norcini, John J and Blank, Linda L and Duffy, F Daniel and Fortna, Gregory S , journal=. The

  85. [93]

    arXiv preprint arXiv:2403.08495 , year=

    Automatic interactive evaluation for large language models with state aware patient simulator , author=. arXiv preprint arXiv:2403.08495 , year=

    arXiv

  86. [94]

    Advances in Neural Information Processing Systems , volume=

    PatientSIM: A persona-driven simulator for realistic doctor-patient interactions , author=. Advances in Neural Information Processing Systems , volume=

  87. [95]

    arXiv preprint arXiv:2503.16547 , year=

    Empowering medical multi-agents with clinical consultation flow for dynamic diagnosis , author=. arXiv preprint arXiv:2503.16547 , year=

    arXiv

  88. [96]

    International Conference on Machine Learning , year=

    Salus: Strategic diagnostic testing for complex diagnosis via multi-agent reinforcement learning , author=. International Conference on Machine Learning , year=

  89. [97]

    arXiv preprint arXiv:2510.24654 , year=

    Evolving diagnostic agents in a virtual clinical environment , author=. arXiv preprint arXiv:2510.24654 , year=

  90. [98]

    2009 , howpublished=

    Conceptual Framework for the International Classification for Patient Safety , author=. 2009 , howpublished=

    2009

  91. [99]

    2022 , howpublished=

    Common Formats for Patient Safety , author=. 2022 , howpublished=

    2022

  92. [100]

    2000 , publisher=

    To Err Is Human: Building a Safer Health System , author=. 2000 , publisher=

    2000

  93. [101]

    Archives of Internal Medicine , volume=

    Diagnostic Error in Medicine: Analysis of 583 Physician-Reported Errors , author=. Archives of Internal Medicine , volume=

  94. [102]

    Chang, Anita and Schyve, Paul M and Croteau, Richard J and others , journal=. The

  95. [103]

    JAMA , volume=

    Physician-Patient Communication: The Relationship with Malpractice Claims among Primary Care Physicians and Surgeons , author=. JAMA , volume=

  96. [104]

    The Health Literacy of

    Kutner, Mark and Greenberg, Elizabeth and Jin, Ying and Paulsen, Christine , journal=. The Health Literacy of. 2006 , publisher=

    2006

  97. [105]

    Journal of General Internal Medicine , volume=

    Prevalence of Limited Health Literacy , author=. Journal of General Internal Medicine , volume=

  98. [106]

    Public Health Reports , volume=

    Race and Trust in the Health Care System , author=. Public Health Reports , volume=

  99. [107]

    Health Services Research , volume=

    Mistrust of Health Care Organizations Is Associated with Underutilization of Health Services , author=. Health Services Research , volume=

  100. [108]

    Medical Care , volume=

    Physician Communication and Patient Adherence to Treatment: A Meta-Analysis , author=. Medical Care , volume=

  101. [109]

    Academic Emergency Medicine , volume=

    Achieving Quality in Clinical Decision Making: Cognitive Strategies and Detection of Bias , author=. Academic Emergency Medicine , volume=

  102. [110]

    Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing

    Red Teaming Language Models with Language Models , author=. Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing. doi:10.18653/v1/2022.emnlp-main.225

    2022 doi

  103. [111]

    arXiv preprint arXiv:2307.15043 , year=

    Universal and Transferable Adversarial Attacks on Aligned Language Models , author=. arXiv preprint arXiv:2307.15043 , year=

    arXiv

  104. [112]

    arXiv preprint arXiv:2209.07858 , year=

    Red Teaming Language Models to Reduce Harms: Methods, Scaling Behaviors, and Lessons Learned , author=. arXiv preprint arXiv:2209.07858 , year=

    arXiv

  105. [113]

    A clinical environment simulator for dynamic

    Luo, Luyang and Kim, Sung Eun and Zhang, Xiaoman and Kernbach, Julius M and Kenia, Roshan and Acosta, Julian N and Nathanson, Larry A and Haimovich, Adrian D and Rodman, Adam and Goh, Ethan and others , journal=. A clinical environment simulator for dynamic. 2026 , publisher=

    2026

  106. [114]

    Gottweis, Juraj and Weng, Wei-Hung and Daryin, Alexander and others , journal=. An

  107. [115]

    Nature Medicine , volume=

    A Toolbox for Surfacing Health Equity Harms and Biases in Large Language Models , author=. Nature Medicine , volume=

  108. [116]

    Nature Medicine , volume=

    Generative Models Improve Fairness of Medical Classifiers under Distribution Shift , author=. Nature Medicine , volume=

  109. [117]

    Nature Medicine , volume=

    Do No Harm: A Roadmap for Responsible Machine Learning for Health Care , author=. Nature Medicine , volume=

  110. [118]

    Nature Medicine , year=

    Evaluation and Mitigation of the Limitations of Large Language Models in Clinical Decision-Making , author=. Nature Medicine , year=

  111. [119]

    Generative

    Teo, Zhen Ling and others , journal=. Generative

  112. [120]

    Scaling medical

    Li, Michelle M and Reis, Ben Y and Rodman, Adam and Cai, Tianxi and Dagan, Noa and Balicer, Ran D and Loscalzo, Joseph and Kohane, Isaac S and Zitnik, Marinka , journal=. Scaling medical. 2026 , publisher=

    2026

  113. [121]

    ACM Computing Surveys , volume=

    Reinforcement Learning in Healthcare: A Survey , author=. ACM Computing Surveys , volume=

  114. [122]

    Nature Medicine , volume=

    Guidelines for Reinforcement Learning in Healthcare , author=. Nature Medicine , volume=

  115. [123]

    Nature Medicine , volume=

    The Artificial Intelligence Clinician Learns Optimal Treatment Strategies for Sepsis in Intensive Care , author=. Nature Medicine , volume=

  116. [124]

    npj Digital Medicine , year=

    Large Language Models in Mental Health Care: A Scoping Review , author=. npj Digital Medicine , year=

  117. [125]

    Academic Medicine , volume=

    An Overview of the Uses of Standardized Patients for Teaching and Evaluating Clinical Skills , author=. Academic Medicine , volume=

  118. [126]

    Assessment of Clinical Competence Using

    Harden, Ronald M and Stevenson, Mary and Downie, Wilson Wilson and Wilson, Gerald M , journal=. Assessment of Clinical Competence Using

  119. [127]

    Fansi Tchango, Arsene and Goel, Rishab and Wen, Zhi and Martel, Julien and Ghosn, Joumana , journal=

  120. [128]

    Johnson, Alistair EW and Pollard, Tom J and Shen, Lu and others , journal=

  121. [129]

    Proceedings of the 36th Annual ACM Symposium on User Interface Software and Technology , year=

    Generative Agents: Interactive Simulacra of Human Behavior , author=. Proceedings of the 36th Annual ACM Symposium on User Interface Software and Technology , year=

  122. [130]

    Li, Yihuai and Hao, Yu and Fang, Lei , journal=

  123. [131]

    Concrete Problems in

    Amodei, Dario and Olah, Chris and Steinhardt, Jacob and Christiano, Paul and Schulman, John and Man. Concrete Problems in. arXiv preprint arXiv:1606.06565 , year=

    arXiv

  124. [132]

    arXiv preprint arXiv:2505.08775 , year=

    Healthbench: Evaluating large language models towards improved human health , author=. arXiv preprint arXiv:2505.08775 , year=

    arXiv

  125. [133]

    Journal of General Internal Medicine , volume=

    Unmet Expectations for Care and the Patient-Physician Relationship , author=. Journal of General Internal Medicine , volume=

  126. [134]

    CMAJ , volume=

    Effective Physician-Patient Communication and Health Outcomes: A Review , author=. CMAJ , volume=

  127. [135]

    Artificial Intelligence , volume=

    Planning and Acting in Partially Observable Stochastic Domains , author=. Artificial Intelligence , volume=. 1998 , publisher=

    1998

  128. [136]

    National Cancer Institute, NIH Publication , year=

    Patient-Centered Communication in Cancer Care: Promoting Healing and Reducing Suffering , author=. National Cancer Institute, NIH Publication , year=

  129. [137]

    2021 , howpublished=

    National Health Interview Survey, 2021 , author=. 2021 , howpublished=

    2021

  130. [138]

    2024 , howpublished=

    2024

  131. [139]

    Zheng, Lianmin and Chiang, Wei-Lin and Sheng, Ying and Zhuang, Siyuan and Wu, Zhanghao and Zhuang, Yonghao and Lin, Zi and Li, Zhuohan and Li, Dacheng and Xing, Eric P and Zhang, Hao and Gonzalez, Joseph E and Stoica, Ion , journal=. Judging

  132. [140]

    Shao, Zhihong and Wang, Peiyi and Zhu, Qihao and Xu, Runxin and Song, Junxiao and Zhang, Mingchuan and Li, Y K and Wu, Y and Guo, Daya , journal=

  133. [141]

    arXiv preprint arXiv:2312.11805 , year=

    Gemini: A Family of Highly Capable Multimodal Models , author=. arXiv preprint arXiv:2312.11805 , year=

    arXiv

  134. [142]

    Gemini 3.1 Pro Model Card , author=

  135. [143]

    Gemini 3.5 Flash Model Card , author=

  136. [144]

    Li, Junhua and Wang, Jingyu and Zhang, Siyu and others , journal=

  137. [145]

    Kyung, Daeun and Chung, Hyunseung and Bae, Seongsu and Kim, Jiho and Sohn, Jae Ho and Kim, Taerim and Kim, Soo Kyung and Choi, Edward , journal=

  138. [146]

    Findings of the Association for Computational Linguistics: ACL 2024 , pages=

    Medagents: Large language models as collaborators for zero-shot medical reasoning , author=. Findings of the Association for Computational Linguistics: ACL 2024 , pages=

    2024

  139. [147]

    Applied Sciences , volume=

    What Disease does this Patient Have? A Large-scale Open Domain Question Answering Dataset from Medical Exams , author=. Applied Sciences , volume=

  140. [148]

    Nature medicine , volume=

    An evaluation framework for clinical use of large language models in patient interaction tasks , author=. Nature medicine , volume=. 2025 , publisher=

    2025

  141. [149]

    Li, Yunxiang and Li, Zihan and Zhang, Kai and Dan, Ruilong and Jiang, Steve and Zhang, You , journal=

  142. [150]

    arXiv preprint arXiv:1707.06347 , year=

    Proximal Policy Optimization Algorithms , author=. arXiv preprint arXiv:1707.06347 , year=

    arXiv

  143. [151]

    Advances in Neural Information Processing Systems , volume=

    Direct Preference Optimization: Your Language Model is Secretly a Reward Model , author=. Advances in Neural Information Processing Systems , volume=

  144. [152]

    Guo, Daya and Yang, Dejian and Zhang, He and others , journal=

  145. [153]

    Nature Medicine , volume=

    High-Performance Medicine: The Convergence of Human and Artificial Intelligence , author=. Nature Medicine , volume=. 2019 , publisher=

    2019

  146. [154]

    2022 , publisher=

    Rajpurkar, Pranav and Chen, Emma and Banerjee, Oishi and Topol, Eric J , journal=. 2022 , publisher=

    2022

  147. [155]

    Nature , volume=

    Foundation Models for Generalist Medical Artificial Intelligence , author=. Nature , volume=. 2023 , publisher=

    2023

  148. [156]

    Nature Medicine , volume=

    Large Language Models in Medicine , author=. Nature Medicine , volume=. 2023 , publisher=

    2023

  149. [157]

    JAMA Internal Medicine , volume=

    Comparing Physician and Artificial Intelligence Chatbot Responses to Patient Questions Posted to a Social Media Forum , author=. JAMA Internal Medicine , volume=

  150. [158]

    Proceedings of the IEEE , volume=

    Young, Steve and Ga. Proceedings of the IEEE , volume=

  151. [159]

    Optimizing Dialogue Management with Reinforcement Learning: Experiments with the

    Singh, Satinder and Litman, Diane and Kearns, Michael and Walker, Marilyn , journal=. Optimizing Dialogue Management with Reinforcement Learning: Experiments with the

  152. [160]

    The Knowledge Engineering Review , volume=

    A Survey of Statistical User Simulation Techniques for Reinforcement-Learning of Dialogue Management Strategies , author=. The Knowledge Engineering Review , volume=

  153. [161]

    Brockman, Greg and Cheung, Vicki and Pettersson, Ludwig and Schneider, Jonas and Schulman, John and Tang, Jie and Zaremba, Wojciech , journal=

  154. [162]

    arXiv preprint arXiv:1909.08593 , year=

    Fine-Tuning Language Models from Human Preferences , author=. arXiv preprint arXiv:1909.08593 , year=

    1909 arXiv

  155. [163]

    Advances in Neural Information Processing Systems , volume=

    Learning to Summarize with Human Feedback , author=. Advances in Neural Information Processing Systems , volume=

  156. [164]

    Proceedings of the Sixteenth International Conference on Machine Learning , pages=

    Policy Invariance Under Reward Transformations: Theory and Application to Reward Shaping , author=. Proceedings of the Sixteenth International Conference on Machine Learning , pages=

  157. [165]

    arXiv preprint arXiv:2305.20050 , year=

    Let's Verify Step by Step , author=. arXiv preprint arXiv:2305.20050 , year=

    arXiv

  158. [166]

    arXiv preprint arXiv:1909.07528 , year=

    Emergent Tool Use from Multi-Agent Autocurricula , author=. arXiv preprint arXiv:1909.07528 , year=

    1909 arXiv

  159. [167]

    International Conference on Learning Representations , year=

    Intrinsic Motivation and Automatic Curricula via Asymmetric Self-Play , author=. International Conference on Learning Representations , year=

  160. [168]

    2015 , publisher=

    Improving Diagnosis in Health Care , author=. 2015 , publisher=

    2015

  161. [169]

    Archives of Internal Medicine , volume=

    Diagnostic Error in Internal Medicine , author=. Archives of Internal Medicine , volume=

  162. [170]

    JAMA Internal Medicine , volume=

    Types and Origins of Diagnostic Errors in Primary Care Settings , author=. JAMA Internal Medicine , volume=

  163. [171]

    npj Digital Medicine , volume=

    Large Language Models Propagate Race-Based Medicine , author=. npj Digital Medicine , volume=

  164. [172]

    arXiv preprint arXiv:2401.10020 , year=

    Self-Rewarding Language Models , author=. arXiv preprint arXiv:2401.10020 , year=

    arXiv

  165. [173]

    Assessing Risks of Large Language Models in Mental Health Support: A Framework for Automated Clinical

    Steenstra, Iain and others , journal=. Assessing Risks of Large Language Models in Mental Health Support: A Framework for Automated Clinical

  166. [174]

    arXiv preprint arXiv:2207.05221 , year=

    Language Models (Mostly) Know What They Know , author=. arXiv preprint arXiv:2207.05221 , year=

    arXiv

  167. [175]

    Hazra, Rishi and de Melo, Pedro and Pinheiro, Tiago and others , journal=

  168. [176]

    arXiv preprint , year=

    Interactive Planning Using Large Language Models for Partially Observable Robotics Tasks , author=. arXiv preprint , year=

  169. [177]

    Fleming, Scott L and Lozano, Alejandro and Haberkorn, William J and Jindal, Jenelle A and Reis, Eduardo and Thapa, Rahul and Blankemeier, Louis and Genkins, Julian Z and Steinberg, Ethan and Nayak, Ashwin and others , booktitle=

  170. [178]

    Liu, Xiao and Yu, Hao and Zhang, Hanchen and Xu, Yifan and Lei, Xuanyu and Lai, Hanyu and Gu, Yu and Ding, Hangliang and Men, Kaiwen and Yang, Kejuan and others , journal=

  171. [179]

    Evaluating

    Guan, Shengyue and Wang, Jindong and Bian, Jiang and Zhu, Bin and Lou, Jian-Guang and Xiong, Haoyi , journal=. Evaluating. 2026 , publisher=

    2026

  172. [180]

    Science , volume=

    Performance of a large language model on the reasoning tasks of a physician , author=. Science , volume=. 2026 , publisher=

    2026

  173. [181]

    Academic Medicine , volume =

    The assessment of clinical skills/competence/performance , author =. Academic Medicine , volume =. 1990 , publisher =

    1990

  174. [182]

    2026 , howpublished =

    2026

  175. [183]

    arXiv preprint arXiv:2504.01848 , year=

    PaperBench: Evaluating AI's Ability to Replicate AI Research , author=. arXiv preprint arXiv:2504.01848 , year=

    arXiv

  176. [184]

    First Conference on Language Modeling , year=

    Automata-based constraints for language model decoding , author=. First Conference on Language Modeling , year=

  177. [185]

    The use of scoring rubrics: Reliability, validity and educational consequences

    Jonsson, Anders and Svingby, Gunilla. The use of scoring rubrics: Reliability, validity and educational consequences. Educ. Res. Rev. doi:10.1016/j.edurev.2007.05.002

    2007 doi

  178. [186]

    A clinical environment simulator for dynamic AI evaluation

    Luo, Luyang and Kim, Sung Eun and Zhang, Xiaoman and Kernbach, Julius M and Kenia, Roshan and Acosta, Julian N and Nathanson, Larry A and Haimovich, Adrian D and Rodman, Adam and Goh, Ethan and Chen, Jonathan H and Shah, Nigam H and Kim, David A and Zou, James and Mahmood, Fai...

    doi

  179. [187]

    Can generalist foundation models outcompete special-purpose tuning? Case study in medicine

    Nori, Harsha and Lee, Yin Tat and Zhang, Sheng and Carignan, Dean and Edgar, Richard and Fusi, Nicolo and King, Nicholas and Larson, Jonathan and Li, Yuanzhi and Liu, Weishung and Luo, Renqian and McKinney, Scott Mayer and Ness, Robert Osazuwa and Poon, Hoifung and Qin, Tao an...

    doi arXiv

  180. [188]

    Genome Medicine , volume=

    Digital twins to personalize medicine , author=. Genome Medicine , volume=. 2019 , publisher=

    2019

  181. [189]

    npj Digital Medicine , volume=

    The health digital twin to tackle cardiovascular disease---a review of an emerging interdisciplinary field , author=. npj Digital Medicine , volume=. 2022 , publisher=

    2022

  182. [190]

    International Conference on Learning Representations (ICLR) , year=

    WebArena: A Realistic Web Environment for Building Autonomous Agents , author=. International Conference on Learning Representations (ICLR) , year=

  183. [191]

    Jimenez, Carlos E and Yang, John and Wettig, Alexander and Yang, Shunyu and Yao, Shunyu and Narasimhan, Karthik and Press, Ofir , booktitle=

  184. [192]

    Multi-agent reinforcement learning: A selective overview of theories and algorithms

    Zhang, Kaiqing and Yang, Zhuoran and Başar, Tamer. Multi-agent reinforcement learning: A selective overview of theories and algorithms. arXiv [cs.LG]. doi:10.48550/arXiv.1911.10635. arXiv:1911.10635

    1911 doi arXiv

  185. [193]

    Assessing electronic note quality using the Physician Documentation Quality Instrument ( PDQI -9)

    Stetson, Peter D and Bakken, Suzanne and Wrenn, Jesse O and Siegler, Eugenia L. Assessing electronic note quality using the Physician Documentation Quality Instrument ( PDQI -9). Appl. Clin. Inform. doi:10.4338/aci-2011-11-ra-0070

    2011 doi

  186. [194]

    Towards conversational medical AI with eyes, ears and a voice

    Shah, Meet and Gusdorf, Jason and Palepu, Anil and Park, Chunjong and O'Sullivan, Jack W and Ravi, Vishnu and Strother, Tim and Dubov, Pavel and Rysbek, Aliya and Fukuzawa, Toshiyuki and Lunts, Yana and Freyberg, Jan and Chang, Michael B and Raghu, Aniruddh and Stutz, David an...

    doi arXiv

  187. [195]

    LLM -as-a-Verifier: A general-purpose verification framework

    Kwok, Jacky and Li, Shulu and Atreya, Pranav and Liu, Yuejiang and Jiang, Yixing and Finn, Chelsea and Pavone, Marco and Stoica, Ion and Mirhoseini, Azalia. LLM -as-a-Verifier: A general-purpose verification framework. arXiv [cs.AI]. doi:10.48550/arXiv.2607.05391. arXiv:2607.05391

    doi arXiv

  188. [196]

    Scaling laws for reward model overoptimization

    Gao, Leo and Schulman, John and Hilton, Jacob. Scaling laws for reward model overoptimization. arXiv [cs.LG]. doi:10.48550/arXiv.2210.10760. arXiv:2210.10760

    doi arXiv

  189. [197]

    Language Models are Few-Shot Learners

    Brown, Tom B and Mann, Benjamin and Ryder, Nick and Subbiah, Melanie and Kaplan, Jared and Dhariwal, Prafulla and Neelakantan, Arvind and Shyam, Pranav and Sastry, Girish and Askell, Amanda and Agarwal, Sandhini and Herbert-Voss, Ariel and Krueger, Gretchen and Henighan, Tom a...

    doi arXiv

  190. [198]

    Polaris: A safety-focused LLM constellation architecture for healthcare

    Mukherjee, Subhabrata and Gamble, Paul and Ausin, Markel Sanz and Kant, Neel and Aggarwal, Kriti and Manjunath, Neha and Datta, Debajyoti and Liu, Zhengliang and Ding, Jiayuan and Busacca, Sophia and Bianco, Cezanne and Sharma, Swapnil and Lasko, Rae and Voisard, Michelle and ...

    arXiv

  191. [199]

    Sequential diagnosis with language models

    Nori, Harsha and Daswani, Mayank and Kelly, Christopher and Lundberg, Scott and Ribeiro, Marco Tulio and Wilson, Marc and Liu, Xiaoxuan and Sounderajah, Viknesh and Carlson, Jonathan and Lungren, Matthew P and Gross, Bay and Hames, Peter and Suleyman, Mustafa and King, Dominic...

    doi arXiv

  192. [200]

    Differential diagnosis generators: an evaluation of currently available computer programs

    Bond, William F and Schwartz, Linda M and Weaver, Kevin R and Levick, Donald and Giuliano, Michael and Graber, Mark L. Differential diagnosis generators: an evaluation of currently available computer programs. J. Gen. Intern. Med. doi:10.1007/s11606-011-1804-8

    doi

  193. [201]

    RubricsTree : Scalable and evolving open-ended evaluation of personal health agents across health memory and medical skills

    Zhang, Weizhi and Li, Zechen and Palangi, Hamid and Graef, Ben and Ali Heydari, A and Lee, Simon A and Rahman, Salman and Luo, Ray and Esmaeilpour, Zeinab and Schenck, Erik and Zhang, Chloe and Li, Yamin and Zhou, Menglian and Yu, Philip S and McDuff, Daniel and Sunden, Lindse...

    doi arXiv

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.