pith. sign in

arxiv: 2606.28526 · v1 · pith:QULH2YVTnew · submitted 2026-06-26 · 💻 cs.CL · cs.HC

A French OSCE Dialogue Dataset and Controllable Virtual Patient System for Clinical Training

Doria Bonzi , Tom Bourgeade , Fabrice Lef\`evre , Irina Illina This is my paper

Pith reviewed 2026-06-30 01:27 UTC · model grok-4.3

classification 💻 cs.CL cs.HC
keywords French OSCE datasetvirtual patientsLLM dialogue generationclinical trainingcontrollable pipelinepatient fidelitymedical education
0
0 comments X

The pith

A French dataset of 240 OSCE interactions supports a controllable LLM pipeline that generates virtual patients with improved fidelity for medical student training.

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

The paper introduces a new dataset of French doctor-patient dialogues from OSCE scenarios to address the shortage of human patients for training medical students. It then builds a modular LLM system that uses retrieval grounding and reflection to control the virtual patient's responses for realism and consistency. Experiments test whether adding these control modules enhances the quality of simulated interactions and the reliability of automated student assessments. If successful, this would allow scalable practice opportunities in French medical education where real standardized patients are limited.

Core claim

The authors present a French OSCE dialogue dataset of 240 interactions and a controllable pipeline that integrates retrieval-based grounding and a reflection loop to generate synthetic dialogues, showing through LLM-as-Judge evaluation that the controllability modules enhance patient fidelity and student evaluation consistency.

What carries the argument

A controllable LLM-based pipeline with retrieval-based grounding and reflection loop for ensuring patient fidelity in virtual OSCE simulations.

If this is right

  • Students gain access to on-demand practice with virtual patients in French.
  • Automated multi-level feedback becomes feasible for communication skills training.
  • The approach can be extended to generate more diverse scenarios.
  • Consistency in evaluations improves across simulated sessions.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • This system could reduce reliance on human actors in medical education globally if adapted to other languages.
  • Integration with real-time speech interfaces might create fully interactive training environments.
  • The dataset could serve as a benchmark for other dialogue generation models in healthcare.

Load-bearing premise

The assumption that an LLM acting as a judge can reliably and accurately assess patient simulation quality and student performance without validation from human experts.

What would settle it

A study where human medical educators rate the same simulations and compare their scores to those from the LLM-as-a-Judge framework, checking for agreement.

Figures

Figures reproduced from arXiv: 2606.28526 by Doria Bonzi, Fabrice Lef\`evre, Irina Illina, Tom Bourgeade.

Figure 1
Figure 1. Figure 1: Overview of the controllable LLM-based dialogue generation pipeline for OSCE simulations [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Overview of the French OSCE dialogue dataset, both recorded and generated. [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: Interactive demo of our VP system. The student interacts with the VP via messages and can consult the physician sheet at any time. An 8-minute timer simulates OSCE conditions. 2. Physician sheet, providing the clin￾ical context and setting, available in￾formations, and consultation objectives (e.g., current medication and information￾gathering goals); 3. Evaluation sheet, listing OSCE criteria the physicia… view at source ↗
Figure 4
Figure 4. Figure 4: Naturalness score across all models and recorded dialogues: bars represent baseline per￾formance averaged over four generated dialogues, deltas indicate performance changes with the re￾trieval and reflection loop for five representative stations. roles. Physician scores are consistently slightly higher than patient simulation scores, sug￾gesting that patient generation may introduce more variability than p… view at source ↗
read the original abstract

The clinical and communication skills of medical students are commonly assessed through Objective Structured Clinical Examinations (OSCEs), which consist of brief scenario-driven simulations of doctor-patient interactions. However, training is often limited by the low availability of human standardized patients, motivating the development of realistic virtual patients (VPs). To address this gap, we introduce a French OSCE dialogue dataset comprising 240 student-patient training interactions. We build upon it a controllable LLM-based pipeline to generate synthetic OSCE dialogues. The pipeline integrates modular components, such as retrieval-based grounding and a reflection loop, to ensure patient fidelity, coherence, and realism. Additionally, we propose a multi-level evaluation framework assessing patient simulation quality, student performance, and linguistic quality, using an LLM-as-a-Judge approach. Experiments suggest that controllability modules generally improve patient fidelity and student evaluation consistency. Finally, we implement an interactive prototype in which students can practice with a VP and receive automatic feedback.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper introduces a French OSCE dialogue dataset of 240 student-patient training interactions and a controllable LLM-based pipeline for generating synthetic dialogues. The pipeline uses modular components such as retrieval-based grounding and a reflection loop to promote patient fidelity, coherence, and realism. It further proposes a multi-level LLM-as-a-Judge evaluation framework to assess patient simulation quality, student performance, and linguistic quality. Experiments indicate that the controllability modules improve patient fidelity and student evaluation consistency, and the work concludes with an interactive prototype allowing students to practice with virtual patients and receive automatic feedback.

Significance. If the central experimental claims hold after proper validation, the dataset and modular pipeline would provide a valuable resource for French-language medical communication training, addressing the scarcity of standardized patients. The controllable generation approach and multi-level evaluation framework could serve as a template for domain-specific dialogue systems. The absence of human-expert calibration for the LLM judge, however, substantially reduces the strength of the reported improvements.

major comments (2)
  1. [Evaluation Framework / Experiments] Evaluation Framework (described in the experiments and evaluation sections): The headline result that controllability modules improve patient fidelity and student evaluation consistency rests entirely on scores from the proposed multi-level LLM-as-a-Judge framework. No correlation, calibration, or agreement statistics with human medical experts are reported on the same criteria, making it impossible to determine whether the observed gains reflect genuine clinical realism or merely LLM-internal biases.
  2. [Dataset] Dataset section: The manuscript states that the dataset comprises 240 interactions but provides no quantitative statistics (e.g., distribution across scenarios, average dialogue length, inter-annotator agreement if any human labeling occurred) or details on collection protocol, participant recruitment, or ethical approval. These omissions are load-bearing for claims about the dataset's utility and the generalizability of the pipeline trained on it.
minor comments (2)
  1. [Abstract / Introduction] The abstract and introduction would benefit from explicit forward references to the specific sections containing dataset statistics, experimental setup, and human-validation results (or their absence).
  2. [Pipeline Description] Notation for the controllability modules (retrieval grounding, reflection loop) should be defined consistently when first introduced and reused in the evaluation tables or figures.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback. We address the two major comments below and indicate the revisions we will make to strengthen the manuscript.

read point-by-point responses

  1. Referee: [Evaluation Framework / Experiments] Evaluation Framework (described in the experiments and evaluation sections): The headline result that controllability modules improve patient fidelity and student evaluation consistency rests entirely on scores from the proposed multi-level LLM-as-a-Judge framework. No correlation, calibration, or agreement statistics with human medical experts are reported on the same criteria, making it impossible to determine whether the observed gains reflect genuine clinical realism or merely LLM-internal biases.

    Authors: We agree that the lack of human-expert calibration or agreement statistics for the LLM-as-Judge is a substantive limitation. The reported improvements in patient fidelity and evaluation consistency are based solely on the LLM judge outputs, and without external validation it is not possible to rule out model-specific biases. In the revised manuscript we will explicitly discuss this limitation in a dedicated subsection and add any available preliminary human agreement data if it can be obtained without new data collection; otherwise we will frame the LLM-as-Judge results as an internal consistency metric rather than a definitive clinical validation. revision: partial

  2. Referee: [Dataset] Dataset section: The manuscript states that the dataset comprises 240 interactions but provides no quantitative statistics (e.g., distribution across scenarios, average dialogue length, inter-annotator agreement if any human labeling occurred) or details on collection protocol, participant recruitment, or ethical approval. These omissions are load-bearing for claims about the dataset's utility and the generalizability of the pipeline trained on it.

    Authors: We acknowledge that the current manuscript lacks the requested quantitative and procedural details. The revised version will expand the Dataset section to include: scenario distribution, average and range of dialogue lengths, any inter-annotator statistics, a full description of the collection protocol, participant recruitment process, and confirmation of ethical approval and consent procedures. revision: yes

Circularity Check

0 steps flagged

No significant circularity; dataset curation and systems work is self-contained

full rationale

The paper introduces a French OSCE dialogue dataset of 240 interactions, constructs a modular LLM pipeline with retrieval grounding and reflection loop for synthetic dialogue generation, and proposes a multi-level LLM-as-a-Judge evaluation framework. No mathematical derivations, fitted parameters renamed as predictions, self-citation load-bearing premises, uniqueness theorems, or ansatzes smuggled via citation are present. The experimental suggestion that controllability modules improve fidelity is an empirical observation from the proposed system and judge, not a reduction to inputs by construction. This matches the default expectation for non-circular systems/dataset papers.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on the domain assumption that 240 collected interactions suffice to ground a generalizable generation pipeline and that LLM-as-a-Judge outputs correlate with human judgments of fidelity and performance; no free parameters or invented entities are introduced.

axioms (1)
  • domain assumption LLM-as-a-Judge provides a valid proxy for human evaluation of patient fidelity and student performance
    Invoked in the multi-level evaluation framework described in the abstract.

pith-pipeline@v0.9.1-grok · 5701 in / 1163 out tokens · 31188 ms · 2026-06-30T01:27:06.844810+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

119 extracted references · 75 canonical work pages · 5 internal anchors

  1. [1]

    Proceedings of the

    Chirkova, Nadezhda and Ajayi, Tunde Oluwaseyi and Aycock, Seth and Mujahid, Zain Muhammad and Perli. Proceedings of the. doi:10.18653/v1/2026.mme-main.7 , url =

    work page doi:10.18653/v1/2026.mme-main.7 2026

  2. [2]

    The Innovation7(6), 101253 (2026)

    Gu, Jiawei and Jiang, Xuhui and Shi, Zhichao and Tan, Hexiang and Zhai, Xuehao and Xu, Chengjin and Li, Wei and Shen, Yinghan and Ma, Shengjie and Liu, Honghao and Wang, Saizhuo and Zhang, Kun and Lin, Zhouchi and Zhang, Bowen and Ni, Lionel and Gao, Wen and Wang, Yuanzhuo and Guo, Jian , year = 2026, month = jan, journal =. A Survey on. doi:10.1016/j.xin...

    work page doi:10.1016/j.xinn.2025.101253 2026

  3. [3]

    Proceedings of the 2020

    Zeng, Guangtao and Yang, Wenmian and Ju, Zeqian and Yang, Yue and Wang, Sicheng and Zhang, Ruisi and Zhou, Meng and Zeng, Jiaqi and Dong, Xiangyu and Zhang, Ruoyu and Fang, Hongchao and Zhu, Penghui and Chen, Shu and Xie, Pengtao , editor =. Proceedings of the 2020. doi:10.18653/v1/2020.emnlp-main.743 , url =

    work page doi:10.18653/v1/2020.emnlp-main.743 2020

  4. [4]

    Proceedings of the 2024

    Labrak, Yanis and Bazoge, Adrien and El Khettari, Oumaima and Rouvier, Mickael and Constant Dit Beaufils, Pacome and Grabar, Natalia and Daille, B. Proceedings of the 2024

    2024

  5. [5]

    Semantic Similarity to Improve Question Understanding in a Virtual Patient , booktitle =

    Laleye, Fr. Semantic Similarity to Improve Question Understanding in a Virtual Patient , booktitle =. doi:10.1145/3341105.3373936 , url =

    work page doi:10.1145/3341105.3373936

  6. [6]

    Li, Tao and Li, Gang and Deng, Zhiwei and Wang, Bryan and Li, Yang , editor =. A. Findings of the. doi:10.18653/v1/2023.findings-emnlp.753 , url =

    work page doi:10.18653/v1/2023.findings-emnlp.753 2023

  7. [7]

    Radford, Alec and Kim, Jong Wook and Xu, Tao and Brockman, Greg and Mcleavey, Christine and Sutskever, Ilya , year = 2023, month = jul, pages =. Robust. Proceedings of the 40th

    2023

  8. [8]

    Proceedings of the 2024

    Saley, Vishal Vivek and Saha, Goonjan and Das, Rocktim Jyoti and Raghu, Dinesh and ., Mausam , editor =. Proceedings of the 2024. doi:10.18653/v1/2024.emnlp-main.936 , url =

    work page doi:10.18653/v1/2024.emnlp-main.936 2024

  9. [10]

    and Stoica, Ion , year = 2023, month = dec, journal =

    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 and Zhang, Hao and Gonzalez, Joseph E. and Stoica, Ion , year = 2023, month = dec, journal =. Judging

    2023

  10. [11]

    International Journal of Clinical Pharmacy , year =

    Yusuf Alzahrani and Sarah Rehman and Ejaz Cheema , title =. International Journal of Clinical Pharmacy , year =. doi:10.1007/s11096-025-01876-5 , url =

    work page doi:10.1007/s11096-025-01876-5

  11. [12]

    2025 , eprint=

    LLM-Powered Virtual Patient Agents for Interactive Clinical Skills Training with Automated Feedback , author=. 2025 , eprint=

    2025

  12. [13]

    Gemini 3.1 Flash-Lite , year =

  13. [14]

    Towards A Deep Learning Question-Answering Specialized Chatbot for Objective Structured Clinical Examinations , year=

    Zini, Julia El and Rizk, Yara and Awad, Mariette and Antoun, Jumana , booktitle=. Towards A Deep Learning Question-Answering Specialized Chatbot for Objective Structured Clinical Examinations , year=

  14. [15]

    Enhancing Clinical Reasoning with Virtual Patients: A Hybrid Systematic Review Combining Human Reviewers and ChatGPT , journal =

    Garc. Enhancing Clinical Reasoning with Virtual Patients: A Hybrid Systematic Review Combining Human Reviewers and ChatGPT , journal =. 2024 , volume =. doi:10.3390/healthcare12222241 , pmid =

    work page doi:10.3390/healthcare12222241 2024

  15. [18]

    and Salisbury, C

    Jepson, M. and Salisbury, C. and Ridd, M. J. and Metcalfe, C. and Garside, L. and Barnes, R. K. , title =. British Journal of General Practice , year =. doi:10.3399/bjgp17x690521 , url =

    work page doi:10.3399/bjgp17x690521

  16. [19]

    NoteChat: A Dataset of Synthetic Patient-Physician Conversations Conditioned on Clinical Notes , url=

    Wang, Junda and Yao, Zonghai and Yang, Zhichao and Zhou, Huixue and Li, Rumeng and Wang, Xun and Xu, Yucheng and Yu, Hong , year=. NoteChat: A Dataset of Synthetic Patient-Physician Conversations Conditioned on Clinical Notes , url=. doi:10.18653/v1/2024.findings-acl.901 , booktitle=

    work page doi:10.18653/v1/2024.findings-acl.901 2024

  17. [20]

    Proceedings of the 17th Conference of the European Chapter of the Association for Computational Linguistics

    Ben Abacha, Asma and Yim, Wen-wai and Fan, Yadan and Lin, Thomas. Proceedings of the 17th Conference of the European Chapter of the Association for Computational Linguistics. 2023

    2023

  18. [21]

    Laleye, Fréjus A. A. and de Chalendar, Gaël and Blanié, Antonia and Brouquet, Antoine and Behnamou, Dan , month = may, year =. A. Proceedings of

  19. [22]

    Favre, Benoit and Cheung, Kyla and Kazemian, Siavash and Lee, Adam and Liu, Yang and Munteanu, Cosmin and Nenkova, Ani and Ochei, Dennis and Penn, Gerald and Tratz, Stephen and Voss, Clare and Zeller, Frauke , year =

  20. [23]

    2021 , eprint=

    Rethinking Evaluation in ASR: Are Our Models Robust Enough? , author=. 2021 , eprint=

    2021

  21. [25]

    Enabling Trait-based Personality Simulation in Conversational LLM Agents: Case Study of Customer Assistance in F rench

    Njifenjou, Ahmed and Sucal, Virgile and Jabaian, Bassam and Lef \`e vre, Fabrice. Enabling Trait-based Personality Simulation in Conversational LLM Agents: Case Study of Customer Assistance in F rench. Proceedings of the 15th International Workshop on Spoken Dialogue Systems Technology. 2025

    2025

  22. [26]

    2024 , eprint=

    Role-Play Zero-Shot Prompting with Large Language Models for Open-Domain Human-Machine Conversation , author=. 2024 , eprint=

    2024

  23. [27]

    2004 , month =

    From WER and RIL to MER and WIL: improved evaluation measures for connected speech recognition , author =. 2004 , month =

    2004

  24. [28]

    M ed D ialog- FR : A F rench Version of the M ed D ialog Corpus for Multi-label Classification and Response Generation Related to Women ' s Intimate Health

    Liu, Xingyu and Segonne, Vincent and Mannion, Aidan and Schwab, Didier and Goeuriot, Lorraine and Portet, Fran c ois. M ed D ialog- FR : A F rench Version of the M ed D ialog Corpus for Multi-label Classification and Response Generation Related to Women ' s Intimate Health. Proceedings of the First Workshop on Patient-Oriented Language Processing (CL4Heal...

    2024

  25. [29]

    2024 , eprint=

    Gemini 1.5: Unlocking multimodal understanding across millions of tokens of context , author=. 2024 , eprint=

    2024

  26. [30]

    Journal of Medical Systems / Elsevier (ScienceDirect) , year=

    Large Language Model–Based Virtual Patient Systems for History-Taking in Medical Education: Comprehensive Systematic Review , author=. Journal of Medical Systems / Elsevier (ScienceDirect) , year=. doi:S2291969426000013 , url=

  27. [31]

    2024 , eprint=

    Synthetic Patient-Physician Dialogue Generation from Clinical Notes Using LLM , author=. 2024 , eprint=

    2024

  28. [32]

    Journal of Medical Internet Research , volume=

    Virtual Patients Using Large Language Models: Scalable, Contextualized Simulation of Clinician-Patient Dialogue With Feedback , author=. Journal of Medical Internet Research , volume=. 2025 , publisher=. doi:10.2196/68486 , url=

    work page doi:10.2196/68486 2025

  29. [33]

    2026 , edition =

    SOAP Notes , author =. 2026 , edition =

    2026

  30. [34]

    Discover Artificial Intelligence , volume=

    Applying state-of-the-art artificial intelligence to grading in simulation-based education: assessment, feedback, and ROI , author=. Discover Artificial Intelligence , volume=. 2025 , doi=

    2025

  31. [35]

    Computational and Structural Biotechnology Journal , volume=

    Integrating large language model-based agents into a virtual patient chatbot for clinical anamnesis training , author=. Computational and Structural Biotechnology Journal , volume=. 2025 , publisher=. doi:10.1016/j.csbj.2025.05.025 , url=

    work page doi:10.1016/j.csbj.2025.05.025 2025

  32. [36]

    arXiv preprint arXiv:2305.13614 , year=

    LLM-empowered Chatbots for Psychiatrist and Patient Simulation: Application and Evaluation , author=. arXiv preprint arXiv:2305.13614 , year=

    arXiv

  33. [37]

    and Zhi, Jiayin and Eack, Shaun M

    Wang, Ruiyi and Milani, Stephanie and Chiu, Jamie C. and Zhi, Jiayin and Eack, Shaun M. and Labrum, Travis and Murphy, Samuel M and Jones, Nev and Hardy, Kate V and Shen, Hong and Fang, Fei and Chen, Zhiyu. PATIENT - : Using Large Language Models to Simulate Patients for Training Mental Health Professionals. Proceedings of the 2024 Conference on Empirical...

    2024

  34. [38]

    2026 , url=

    Lakshya A Agrawal and Shangyin Tan and Dilara Soylu and Noah Ziems and Rishi Khare and Krista Opsahl-Ong and Arnav Singhvi and Herumb Shandilya and Michael J Ryan and Meng Jiang and Christopher Potts and Koushik Sen and Alex Dimakis and Ion Stoica and Dan Klein and Matei Zaharia and Omar Khattab , booktitle=. 2026 , url=

    2026

  35. [39]

    2025 , eprint=

    CUB: Benchmarking Context Utilisation Techniques for Language Models , author=. 2025 , eprint=

    2025

  36. [40]

    2024 , eprint=

    Large Language Models for Medical OSCE Assessment: A Novel Approach to Transcript Analysis , author=. 2024 , eprint=

    2024

  37. [41]

    2024 , eprint=

    GPT-4 Technical Report , author=. 2024 , eprint=

    2024

  38. [42]

    2025 , url =

    Claude Haiku 4.5 System Card , author =. 2025 , url =

    2025

  39. [43]

    arXiv preprint arXiv:2601.08584 , year=

    Ministral 3 , author=. arXiv preprint arXiv:2601.08584 , year=

    Pith/arXiv arXiv

  40. [44]

    Proceedings of the 15th International Conference on Language Resources and Evaluation (LREC 2026) , year =

    LLM-Based Data Generation and Clinical Skills Evaluation for Low-Resource French OSCEs , author =. Proceedings of the 15th International Conference on Language Resources and Evaluation (LREC 2026) , year =

    2026

  41. [45]

    2024 , eprint=

    Biomedical Large Languages Models Seem not to be Superior to Generalist Models on Unseen Medical Data , author=. 2024 , eprint=

    2024

  42. [46]

    Journal of Chiropractic Medicine , volume=

    A guideline of selecting and reporting intraclass correlation coefficients for reliability research , author=. Journal of Chiropractic Medicine , volume=. 2016 , doi=

    2016

  43. [47]

    Frontiers in Virtual Reality , volume=

    Embodied virtual patients as a simulation-based framework for training clinician-patient communication skills: An overview of their use in psychiatric and geriatric care , author=. Frontiers in Virtual Reality , volume=. 2022 , publisher=

    2022

  44. [48]

    fr\_core\_news\_md: spaCy French model , year =

  45. [49]

    2024 , url =

    Jitsi , title =. 2024 , url =

    2024

  46. [50]

    2024 , url =

    PyAnnoteAI , title =. 2024 , url =

    2024

  47. [51]

    2024 , url =

    OpenAI , title =. 2024 , url =

    2024

  48. [52]

    Advances in Simulation , volume =

    The Association of Standardized Patient Educators (ASPE) Standards of Best Practice , author =. Advances in Simulation , volume =. 2017 , doi =

    2017

  49. [53]

    McNeill, Matthew and Levitan, Rivka , month = aug, year =. An. doi:10.21437/Interspeech.2023-2525 , abstract =

    work page doi:10.21437/interspeech.2023-2525 2023

  50. [55]

    and Cortes, Ariel and Pomares, Alexandra and Delgadillo, Vivian and Yepes, Francisco J

    Gorbanev, Iouri and Agudelo-Londoño, Sandra and González, Rafael A. and Cortes, Ariel and Pomares, Alexandra and Delgadillo, Vivian and Yepes, Francisco J. and Muñoz, Óscar , month = jan, year =. A systematic review of serious games in medical education: quality of evidence and pedagogical strategy , copyright =. Medical Education Online , publisher =

  51. [56]

    JMIR Serious Games , month = jan, year =

    Developing. JMIR Serious Games , month = jan, year =. doi:10.2196/11565 , abstract =

    work page doi:10.2196/11565

  52. [57]

    Huang, Yining and Tang, Keke and Chen, Meilian and Wang, Boyuan , month = may, year =. A. doi:10.48550/arXiv.2404.15777 , abstract =

    work page doi:10.48550/arxiv.2404.15777

  53. [58]

    Ye, Qinyuan and Axmed, Maxamed and Pryzant, Reid and Khani, Fereshte , month = jul, year =. Prompt. doi:10.48550/arXiv.2311.05661 , abstract =

    work page doi:10.48550/arxiv.2311.05661

  54. [59]

    Li, Wenwu and Wang, Xiangfeng and Li, Wenhao and Jin, Bo , month = feb, year =. A. doi:10.48550/arXiv.2502.11560 , abstract =

    work page doi:10.48550/arxiv.2502.11560

  55. [60]

    GEPA: Reflective Prompt Evolution Can Outperform Reinforcement Learning

    Agrawal, Lakshya A. and Tan, Shangyin and Soylu, Dilara and Ziems, Noah and Khare, Rishi and Opsahl-Ong, Krista and Singhvi, Arnav and Shandilya, Herumb and Ryan, Michael J. and Jiang, Meng and Potts, Christopher and Sen, Koushik and Dimakis, Alexandros G. and Stoica, Ion and Klein, Dan and Zaharia, Matei and Khattab, Omar , month = feb, year =. doi:10.48...

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2507.19457

  56. [61]

    doi:10.48550/arXiv.2506.09147 , abstract =

    Chirkova, Nadezhda and Ajayi, Tunde Oluwaseyi and Aycock, Seth and Mujahid, Zain Muhammad and Perlić, Vladana and Borisova, Ekaterina and Vartampetian, Markarit , month = dec, year =. doi:10.48550/arXiv.2506.09147 , abstract =

    work page doi:10.48550/arxiv.2506.09147

  57. [62]

    Li, Tao and Li, Gang and Deng, Zhiwei and Wang, Bryan and Li, Yang , month = oct, year =. A. doi:10.48550/arXiv.2310.08740 , abstract =

    work page doi:10.48550/arxiv.2310.08740

  58. [63]

    Chain-of-Thought Prompting Elicits Reasoning in Large Language Models

    Wei, Jason and Wang, Xuezhi and Schuurmans, Dale and Bosma, Maarten and Ichter, Brian and Xia, Fei and Chi, Ed and Le, Quoc and Zhou, Denny , month = jan, year =. Chain-of-. doi:10.48550/arXiv.2201.11903 , abstract =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2201.11903

  59. [64]

    Singhal, Karan and Azizi, Shekoofeh and Tu, Tao and Mahdavi, S. Sara and Wei, Jason and Chung, Hyung Won and Scales, Nathan and Tanwani, Ajay and Cole-Lewis, Heather and Pfohl, Stephen and Payne, Perry and Seneviratne, Martin and Gamble, Paul and Kelly, Chris and Scharli, Nathaneal and Chowdhery, Aakanksha and Mansfield, Philip and Arcas, Blaise Aguera y ...

    work page doi:10.48550/arxiv.2212.13138

  60. [65]

    and Xiong, Caiming and Joty, Shafiq and Wu, Chien-Sheng , month = may, year =

    Laban, Philippe and Kryściński, Wojciech and Agarwal, Divyansh and Fabbri, Alexander R. and Xiong, Caiming and Joty, Shafiq and Wu, Chien-Sheng , month = may, year =. doi:10.48550/arXiv.2305.14540 , abstract =

    work page doi:10.48550/arxiv.2305.14540

  61. [66]

    Gu, Jiawei and Jiang, Xuhui and Shi, Zhichao and Tan, Hexiang and Zhai, Xuehao and Xu, Chengjin and Li, Wei and Shen, Yinghan and Ma, Shengjie and Liu, Honghao and Wang, Saizhuo and Zhang, Kun and Wang, Yuanzhuo and Gao, Wen and Ni, Lionel and Guo, Jian , month = mar, year =. A. doi:10.48550/arXiv.2411.15594 , abstract =

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2411.15594

  62. [67]

    Lu, Christina and Gallagher, Jack and Michala, Jonathan and Fish, Kyle and Lindsey, Jack , month = jan, year =. The. doi:10.48550/arXiv.2601.10387 , abstract =

    work page doi:10.48550/arxiv.2601.10387

  63. [68]

    Lee, Seungbeen and Lim, Seungwon and Han, Seungju and Oh, Giyeong and Chae, Hyungjoo and Chung, Jiwan and Kim, Minju and Kwak, Beong-woo and Lee, Yeonsoo and Lee, Dongha and Yeo, Jinyoung and Yu, Youngjae , month = mar, year =. Do. doi:10.48550/arXiv.2406.14703 , abstract =

    work page doi:10.48550/arxiv.2406.14703

  64. [69]

    Neural Networks , author =

    Learning emotions latent representation with. Neural Networks , author =. 2021 , keywords =. doi:10.1016/j.neunet.2021.04.021 , abstract =

    work page doi:10.1016/j.neunet.2021.04.021 2021

  65. [70]

    and Ward, Andy and Ahmed, Sarah and Tanna, Vinay and Rahman, Ismail and Bostock, Jennifer and Nockels, Keith and Wang, Wen and Gillies, Clare L

    Keshtkar, Leila and Madigan, Claire D. and Ward, Andy and Ahmed, Sarah and Tanna, Vinay and Rahman, Ismail and Bostock, Jennifer and Nockels, Keith and Wang, Wen and Gillies, Clare L. and Howick, Jeremy , month = feb, year =. The. Annals of Internal Medicine , publisher =. doi:10.7326/M23-2168 , abstract =

    work page doi:10.7326/m23-2168

  66. [71]

    Sutcliffe, Richard , month = dec, year =. A. doi:10.48550/arXiv.2401.00609 , abstract =

    work page doi:10.48550/arxiv.2401.00609

  67. [72]

    The British Journal of General Practice: The Journal of the Royal College of General Practitioners , author =

    Effectiveness of empathy in general practice: a systematic review , volume =. The British Journal of General Practice: The Journal of the Royal College of General Practitioners , author =. 2013 , keywords =. doi:10.3399/bjgp13X660814 , abstract =

    work page doi:10.3399/bjgp13x660814 2013

  68. [73]

    Manipulation and the

    Krook, Joshua , month = feb, year =. Manipulation and the. doi:10.2139/ssrn.4719835 , abstract =

    work page doi:10.2139/ssrn.4719835

  69. [74]

    Natural Language Engineering , author =

    Designing a virtual patient dialogue system based on terminology-rich resources:. Natural Language Engineering , author =. 2020 , pages =. doi:10.1017/S1351324919000329 , abstract =

    work page doi:10.1017/s1351324919000329 2020

  70. [75]

    IEEE Transactions on Cognitive and Developmental Systems , author =

    Dialogue. IEEE Transactions on Cognitive and Developmental Systems , author =. 2022 , pages =. doi:10.1109/TCDS.2021.3086565 , abstract =

    work page doi:10.1109/tcds.2021.3086565 2022

  71. [76]

    Mao, Shengyu and Wang, Xiaohan and Wang, Mengru and Jiang, Yong and Xie, Pengjun and Huang, Fei and Zhang, Ningyu , month = sep, year =. Editing. doi:10.48550/arXiv.2310.02168 , abstract =

    work page doi:10.48550/arxiv.2310.02168

  72. [77]

    Annales Médico-psychologiques, revue psychiatrique , author =

    Histoire des «. Annales Médico-psychologiques, revue psychiatrique , author =. 2010 , pages =. doi:10.1016/j.amp.2009.04.016 , abstract =

    work page doi:10.1016/j.amp.2009.04.016 2010

  73. [78]

    Advances in

    Personality engineering:. Advances in. 1995 , pages =. doi:10.1016/S0921-2647(06)80262-5 , abstract =

    work page doi:10.1016/s0921-2647(06)80262-5 1995

  74. [79]

    Galland, Lucie , month = oct, year =

  75. [80]

    Kerz, E., Qiao, Y ., Zanwar, S., and Wiechmann, D

    Jiang, Hang and Zhang, Xiajie and Cao, Xubo and Breazeal, Cynthia and Roy, Deb and Kabbara, Jad , month = apr, year =. doi:10.48550/arXiv.2305.02547 , abstract =

    work page doi:10.48550/arxiv.2305.02547

  76. [81]

    Proceedings of the AAAI Conference on Artificial Intelligence , author =

    Personalized. Proceedings of the AAAI Conference on Artificial Intelligence , author =. 2023 , note =. doi:10.1609/aaai.v37i11.26518 , abstract =

    work page doi:10.1609/aaai.v37i11.26518 2023

  77. [82]

    Tailored

    Galland, Lucie and Pelachaud, Catherine and Pecune, Florian , month = jul, year =. Tailored. doi:10.48550/arXiv.2506.19652 , abstract =

    work page doi:10.48550/arxiv.2506.19652

  78. [83]

    Fire, Michael and Elbazis, Yitzhak and Wasenstein, Adi and Rokach, Lior , month = may, year =. Dark. doi:10.48550/arXiv.2505.10066 , abstract =

    work page doi:10.48550/arxiv.2505.10066

  79. [84]

    Computers in Human Behavior , author =

    Exploring relationship development with social chatbots:. Computers in Human Behavior , author =. 2023 , keywords =. doi:10.1016/j.chb.2022.107600 , abstract =

    work page doi:10.1016/j.chb.2022.107600 2023

  80. [85]

    Kovacevic, Nikola and Holz, Christian and Gross, Markus and Wampfler, Rafael , month = oct, year =. A. Proceedings of the 25th. doi:10.1145/3717511.3747070 , abstract =

    work page doi:10.1145/3717511.3747070

Showing first 80 references.