pith. sign in

arxiv: 2605.17765 · v1 · pith:N46XVJAUnew · submitted 2026-05-18 · 💻 cs.LG

AURORA: Contextual Orthogonalization for Geometric Representation Learning in Healthcare Foundation Models

Yuanyun Zhang , Shi Li This is my paper

Pith reviewed 2026-05-20 12:01 UTC · model grok-4.3

classification 💻 cs.LG
keywords healthcare foundation modelsrepresentation disentanglementorthogonal subspacescontextual factorslatent geometryrelational consistencydistribution shift
0
0 comments X

The pith

AURORA decomposes representations in healthcare foundation models into orthogonal semantic subspaces for distinct contextual factors.

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

Healthcare foundation models often mix factors like physiologic severity, intervention intensity, observational structure, and institutional workflow into shared embedding directions. This mixing produces representations that work for prediction but remain hard to interpret and unstable when contexts shift. AURORA addresses the issue by splitting representations into orthogonal subspaces, each tied to one contextual factor. Within each subspace the method enforces relational consistency objectives to preserve geometric structure. The resulting spaces show improved disentanglement and robustness on clinical tasks compared with standard self-supervised approaches.

Core claim

Rather than optimizing a single unified embedding manifold, AURORA decomposes representations into orthogonal semantic subspaces corresponding to distinct contextual factors and learns relational consistency objectives within each subspace. This induces latent spaces that are both semantically disentangled and geometrically interpretable. Across clinical prediction and retrieval tasks the method outperforms reconstruction, contrastive, and self-distillation baselines while raising contextual disentanglement, neighborhood purity, and robustness under institutional distribution shift.

What carries the argument

Contextual orthogonalization, which maps distinct contextual factors to separate orthogonal directions in latent space and applies relational consistency learning inside each subspace.

If this is right

  • The approach yields higher performance than reconstruction, contrastive, and self-distillation baselines on clinical prediction and retrieval tasks.
  • It raises measures of contextual disentanglement, neighborhood purity, and robustness to institutional distribution shift.
  • Latent geometry becomes an explicit design axis for healthcare foundation models alongside predictive compression objectives.
  • Structuring representation space by contextual semantics supplies a complementary route to conventional self-supervised learning.

Where Pith is reading between the lines

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

  • Similar orthogonalization could separate entangled factors in non-healthcare domains such as vision or language models that mix style and content.
  • The learned subspaces might support selective editing of model behavior for one context while leaving others unchanged.
  • Automated discovery of contextual factors could replace manual mapping and broaden applicability to new datasets.

Load-bearing premise

Distinct contextual factors such as physiologic severity, intervention intensity, observational structure, and institutional workflow can be reliably identified and mapped to orthogonal directions without substantial loss of predictive information.

What would settle it

An experiment showing that enforcing the orthogonal split either causes significant overlap between subspaces or reduces accuracy on downstream clinical prediction tasks would falsify the claimed benefit.

Figures

Figures reproduced from arXiv: 2605.17765 by Shi Li, Yuanyun Zhang.

Figure 1
Figure 1. Figure 1: Overview of AURORA. The proposed framework decomposes patient representations into orthogonal semantic subspaces corresponding to physiological state, intervention structure, observation behavior, and contextual residuals, jointly optimizing contextual alignment and latent orthogonality to produce disentangled and semantically coherent healthcare foundation model repre￾sentations. than supervising these fa… view at source ↗
read the original abstract

Recent healthcare foundation models have achieved strong predictive performance through large scale self supervised learning, yet their latent representations frequently entangle physiologic severity, intervention intensity, observational structure, and institutional workflow into shared embedding directions. While effective for downstream prediction, such representations remain semantically opaque and unstable under contextual shift. We introduce AURORA, Adaptive Uncertainty aware Representations through Orthogonalized Relational Alignment, a new framework for healthcare representation learning based on contextual latent geometry. Rather than optimizing a single unified embedding manifold, AURORA decomposes representations into orthogonal semantic subspaces corresponding to distinct contextual factors and learns relational consistency objectives within each subspace. This induces latent spaces that are both semantically disentangled and geometrically interpretable. Across multiple clinical prediction and retrieval tasks, AURORA consistently outperforms reconstruction, contrastive, and self distillation baselines while substantially improving contextual disentanglement, neighborhood purity, and robustness under institutional distribution shift. Our results suggest that latent geometry itself constitutes an important axis of healthcare foundation model design and that explicitly structuring representation space according to contextual semantics provides a complementary direction beyond conventional predictive compression objectives.

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 / 1 minor

Summary. The paper introduces AURORA (Adaptive Uncertainty aware Representations through Orthogonalized Relational Alignment), a framework for healthcare foundation models. It decomposes latent representations into orthogonal semantic subspaces aligned with distinct contextual factors (physiologic severity, intervention intensity, observational structure, institutional workflow) and optimizes relational consistency objectives within each subspace. The central claim is that this produces semantically disentangled, geometrically interpretable representations that outperform reconstruction, contrastive, and self-distillation baselines on clinical prediction and retrieval tasks while improving disentanglement, neighborhood purity, and robustness to institutional shift.

Significance. If the empirical claims hold with rigorous controls, the work would be significant for healthcare representation learning. It shifts focus from pure predictive compression to explicit structuring of latent geometry according to contextual semantics, addressing entanglement and shift robustness that are persistent issues in clinical foundation models. The approach offers a complementary axis to standard SSL objectives and could inform more interpretable model design.

major comments (2)
  1. [Abstract] Abstract: The central claim of consistent outperformance and improved disentanglement is stated without any quantitative results, error bars, dataset sizes, ablation studies, or statistical tests. This prevents direct verification of whether the reported gains are load-bearing or merely incremental.
  2. [Abstract (and presumed Methods)] The manuscript does not address the risk that enforcing strict orthogonality on statistically dependent clinical factors (e.g., severity directly influencing intervention choice) discards shared predictive variance. No completeness argument, information-loss bound, or ablation quantifying downstream task degradation after orthogonalization is provided, which is load-bearing for the claim that subspaces remain complete for prediction.
minor comments (1)
  1. [Abstract] Abstract: 'self supervised' should be hyphenated as 'self-supervised'.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive and detailed feedback. We appreciate the emphasis on strengthening the abstract with quantitative support and on addressing potential information loss from orthogonality constraints. We respond to each major comment below and indicate the revisions made to the manuscript.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The central claim of consistent outperformance and improved disentanglement is stated without any quantitative results, error bars, dataset sizes, ablation studies, or statistical tests. This prevents direct verification of whether the reported gains are load-bearing or merely incremental.

    Authors: We agree that the abstract, as a high-level summary, would benefit from including key quantitative indicators to allow readers to assess the magnitude of the reported improvements. In the revised manuscript we have updated the abstract to reference specific performance gains (with standard errors), the scale of the evaluation (number of datasets and tasks), and the presence of ablations and statistical testing, while keeping the abstract concise. Full tables, error bars, dataset descriptions, and significance tests remain in the main text and supplementary material. revision: yes

  2. Referee: [Abstract (and presumed Methods)] The manuscript does not address the risk that enforcing strict orthogonality on statistically dependent clinical factors (e.g., severity directly influencing intervention choice) discards shared predictive variance. No completeness argument, information-loss bound, or ablation quantifying downstream task degradation after orthogonalization is provided, which is load-bearing for the claim that subspaces remain complete for prediction.

    Authors: We acknowledge this is an important theoretical and practical concern. While our relational alignment objectives are intended to retain predictive signal across subspaces, we recognize that an explicit completeness argument and empirical quantification were not sufficiently highlighted. We have added a dedicated paragraph in the Methods section deriving a bound on preserved mutual information under the orthogonalization operator and included a new ablation that measures downstream task performance with and without the orthogonality constraint, showing that any loss in predictive variance is small relative to gains in robustness and disentanglement. revision: yes

Circularity Check

0 steps flagged

No circularity: derivation remains self-contained without reduction to fitted inputs

full rationale

The abstract introduces AURORA as a framework that decomposes representations into orthogonal semantic subspaces and applies relational consistency objectives within each. No equations, self-citations, or fitted-parameter renamings are provided that would make any claimed prediction or disentanglement result equivalent to its own inputs by construction. The reader's note explicitly states that no explicit reduction of claimed improvements to quantities already fitted inside the paper is visible. The central claims rest on the design of the orthogonalization and consistency objectives rather than on any self-referential loop or imported uniqueness theorem, making the derivation independent of the patterns that would trigger a positive circularity finding.

Axiom & Free-Parameter Ledger

1 free parameters · 1 axioms · 1 invented entities

The approach rests on the domain assumption that contextual factors are separable into orthogonal directions and that enforcing relational consistency inside each subspace yields both interpretability and robustness gains; no explicit free parameters or invented entities are named in the abstract, but the orthogonalization strength and subspace alignment weights are likely fitted hyperparameters.

free parameters (1)
  • Orthogonalization strength
    Hyperparameter controlling how strictly orthogonality is enforced between subspaces; must be chosen or tuned to balance disentanglement against predictive performance.
axioms (1)
  • domain assumption Distinct contextual factors (severity, intervention intensity, observational structure, institutional workflow) exist and can be aligned to orthogonal latent directions.
    Invoked when the paper states that representations are decomposed into orthogonal semantic subspaces corresponding to these factors.
invented entities (1)
  • Orthogonal semantic subspaces no independent evidence
    purpose: To isolate and make interpretable each contextual factor inside the overall embedding.
    Newly postulated geometric construct introduced by the AURORA framework; no independent falsifiable evidence outside the paper is mentioned in the abstract.

pith-pipeline@v0.9.0 · 5713 in / 1641 out tokens · 59339 ms · 2026-05-20T12:01:46.917011+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

36 extracted references · 36 canonical work pages · 5 internal anchors

  1. [1]

    Foundation models in healthcare: Opportunities, risks & strategies forward

    Anja Thieme, Aditya Nori, Marzyeh Ghassemi, Rishi Bommasani, Tariq Osman Andersen, and Ewa Luger. Foundation models in healthcare: Opportunities, risks & strategies forward. InExtended abstracts of the 2023 CHI conference on human factors in computing systems, pages 1–4,

    work page 2023

  2. [2]

    Foundation models for electronic health records: representation dynamics and transferability.arXiv preprint arXiv:2504.10422,

    Michael C Burkhart, Bashar Ramadan, Zewei Liao, Kaveri Chhikara, Juan C Rojas, William F Parker, and Brett K Beaulieu-Jones. Foundation models for electronic health records: representation dynamics and transferability.arXiv preprint arXiv:2504.10422,

    work page arXiv

  3. [3]

    A foundation model for intensive care: Unlocking generalization across tasks and domains at scale.medRxiv, pages 2025–07,

    Manuel Burger, Daphné Chopard, Malte Londschien, Fedor Sergeev, Hugo Yèche, Rita Kuznetsova, Martin Faltys, Eike Gerdes, Polina Leshetkina, Peter Bühlmann, et al. A foundation model for intensive care: Unlocking generalization across tasks and domains at scale.medRxiv, pages 2025–07,

    work page 2025

  4. [4]

    Bert: Pre-training of deep bidirectional transformers for language understanding

    Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. Bert: Pre-training of deep bidirectional transformers for language understanding. InProceedings of the 2019 conference of 10 the North American chapter of the association for computational linguistics: human language technologies, volume 1 (long and short papers), pages 4171–4186,

    work page 2019

  5. [5]

    Deep Residual Learning for Image Recognition

    URLhttps://arxiv.org/abs/1512.03385. Mingyi He, Bo Li, and Huahui Chen. Multi-scale 3d deep convolutional neural network for hyper- spectral image classification. In2017 IEEE International Conference on Image Processing (ICIP), pages 3904–3908. IEEE,

    work page internal anchor Pith review Pith/arXiv arXiv

  6. [6]

    Yuanyun Zhang and Shi Li

    arXiv preprint arXiv:2503.05768, 2025a. Yuanyun Zhang and Shi Li. Chronoformer: Time-aware transformer architectures for structured clinical event modeling.arXiv preprint arXiv:2504.07373, 2025b. Wu Hao Ran, Xi Xi, Furong Li, Jingyi Lu, Jian Jiang, Hui Huang, Yuzhuan Zhang, and Shi Li. Structured semantics from unstructured notes: Language model approache...

    work page arXiv

  7. [7]

    Can large language models abstract medical coded language? arXiv preprint arXiv:2403.10822,

    Simon A Lee and Timothy Lindsey. Can large language models abstract medical coded language? arXiv preprint arXiv:2403.10822,

    work page arXiv

  8. [8]

    Serialized ehr make for good text representations.arXiv preprint arXiv:2510.13843,

    Zhirong Chou, Quan Qin, and Shi Li. Serialized ehr make for good text representations.arXiv preprint arXiv:2510.13843,

    work page arXiv

  9. [9]

    Learning clinical representations under systematic distribution shift

    Yuanyun Zhang and Shi Li. Learning clinical representations under systematic distribution shift. arXiv preprint arXiv:2603.07348,

    work page arXiv

  10. [10]

    Discriminative representation learning for clinical prediction.arXiv preprint arXiv:2603.20921,

    Yang Zhang, Li Fan, Samuel Lawrence, and Shi Li. Discriminative representation learning for clinical prediction.arXiv preprint arXiv:2603.20921,

    work page arXiv

  11. [11]

    Himae: Hierarchical masked autoencoders discover resolution-specific structure in wearable time series.arXiv preprint arXiv:2510.25785, October

    Simon A Lee, Cyrus Tanade, Hao Zhou, Juhyeon Lee, Megha Thukral, Minji Han, Rachel Choi, Md Sazzad Hissain Khan, Baiying Lu, Migyeong Gwak, et al. Himae: Hierarchical masked autoencoders discover resolution-specific structure in wearable time series.arXiv preprint arXiv:2510.25785, October

    work page arXiv

  12. [12]

    Miller, and Ian Shapiro

    Salar Abbaspourazad, Anshuman Mishra, Joseph Futoma, Andrew C Miller, and Ian Shapiro. Wear- able accelerometer foundation models for health via knowledge distillation.arXiv preprint arXiv:2412.11276,

    work page arXiv

  13. [13]

    Large-scale training of foundation models for wearable biosignals.arXiv preprint arXiv:2312.05409,

    Salar Abbaspourazad, Oussama Elachqar, Andrew C Miller, Saba Emrani, Udhyakumar Nallasamy, and Ian Shapiro. Large-scale training of foundation models for wearable biosignals.arXiv preprint arXiv:2312.05409,

    work page arXiv

  14. [14]

    Brandon Westover, and Jimeng Sun

    Chaoqi Yang, M. Brandon Westover, and Jimeng Sun. Biot: Biosignal transformer for cross-data learning in the wild. InNeurIPS 2023,

    work page 2023

  15. [15]

    11 Megha Thukral, Cyrus Tanade, Simon A Lee, Juhyeon Lee, Hao Zhou, Keum San Chun, Migyeong Gwak, Viswam Nathan, Md Mahbubur Rahman, Li Zhu, et al

    URL https://openreview.net/forum?id= c2LZyTyddi. 11 Megha Thukral, Cyrus Tanade, Simon A Lee, Juhyeon Lee, Hao Zhou, Keum San Chun, Migyeong Gwak, Viswam Nathan, Md Mahbubur Rahman, Li Zhu, et al. Wavelet-driven masked multiscale reconstruction for ppg foundation models.arXiv preprint arXiv:2601.12215,

    work page arXiv

  16. [16]

    Physiology-Aware Masked Cross-Modal Reconstruction for Biosignal Representation Learning

    Hao Zhou, Simon A Lee, Cyrus Tanade, Keum San Chun, Juhyeon Lee, Migyeong Gwak, Megha Thukral, Justin Sung, Eugene Hwang, Mehrab Bin Morshed, et al. Physiology-aware masked cross-modal reconstruction for biosignal representation learning.arXiv preprint arXiv:2605.00973,

    work page internal anchor Pith review Pith/arXiv arXiv

  17. [17]

    Language models are few-shot learners.Advances in neural information processing systems, 33:1877–1901,

    Tom Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared D Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, et al. Language models are few-shot learners.Advances in neural information processing systems, 33:1877–1901,

    work page 1901

  18. [18]

    Text serialization and their relationship with the conventional paradigms of tabular machine learning.arXiv preprint arXiv:2406.13846,

    Kyoka Ono and Simon A Lee. Text serialization and their relationship with the conventional paradigms of tabular machine learning.arXiv preprint arXiv:2406.13846,

    work page arXiv

  19. [19]

    Ehrmamba: Towards generalizable and scalable foundation models for electronic health records

    Adibvafa Fallahpour, Mahshid Alinoori, Wenqian Ye, Xu Cao, Arash Afkanpour, and Amrit Krishnan. Ehrmamba: Towards generalizable and scalable foundation models for electronic health records. arXiv preprint arXiv:2405.14567,

    work page arXiv

  20. [20]

    Zilin Jing, Vincent Jeanselme, Yuta Kobayashi, Simon A Lee, Chao Pang, Aparajita Kashyap, Yanwei Li, Xinzhuo Jiang, and Shalmali Joshi

    URLhttps://openreview.net/forum?id=NialiwI2V6. Zilin Jing, Vincent Jeanselme, Yuta Kobayashi, Simon A Lee, Chao Pang, Aparajita Kashyap, Yanwei Li, Xinzhuo Jiang, and Shalmali Joshi. One loss to rule them all: Marked time-to-event for structured ehr foundation models.arXiv preprint arXiv:2602.00541,

    work page arXiv

  21. [21]

    Core-behrt: A carefully optimized and rigorously evaluated behrt

    Mikkel Odgaard, Kiril Vadimovic Klein, Sanne Møller Thysen, Espen Jimenez-Solem, Martin Sillesen, and Mads Nielsen. Core-behrt: A carefully optimized and rigorously evaluated behrt. arXiv preprint arXiv:2404.15201,

    work page arXiv

  22. [22]

    URLhttps://arxiv.org/abs/1910.10699

    doi: 10.48550/arxiv.1910.10699. URLhttps://arxiv.org/abs/1910.10699. Timo Bertram, Johannes Fürnkranz, and Martin Müller. Contrastive learning of preferences with a contextual infonce loss,

    work page doi:10.48550/arxiv.1910.10699 1910

  23. [23]

    Michael Wornow, Suhana Bedi, Miguel Angel Fuentes Hernandez, Ethan Steinberg, Jason Alan Fries, Christopher Ré, Sanmi Koyejo, and Nigam H Shah

    URLhttps://arxiv.org/abs/2407.05898. Michael Wornow, Suhana Bedi, Miguel Angel Fuentes Hernandez, Ethan Steinberg, Jason Alan Fries, Christopher Ré, Sanmi Koyejo, and Nigam H Shah. Context clues: Evaluating long context models for clinical prediction tasks on ehrs.arXiv preprint arXiv:2412.16178,

    work page arXiv

  24. [24]

    DINOv2: Learning Robust Visual Features without Supervision

    Maxime Oquab, Timothée Darcet, Théo Moutakanni, Huy V o, Marc Szafraniec, Vasil Khalidov, Pierre Fernandez, Daniel Haziza, Francisco Massa, Alaaeldin El-Nouby, et al. Dinov2: Learning robust visual features without supervision.arXiv preprint arXiv:2304.07193,

    work page internal anchor Pith review Pith/arXiv arXiv

  25. [25]

    Hybridna: A hybrid transformer-mamba2 long-range dna language model.arXiv preprint arXiv:2502.10807,

    Mingqian Ma, Guoqing Liu, Chuan Cao, Pan Deng, Tri Dao, Albert Gu, Peiran Jin, Zhao Yang, Yingce Xia, Renqian Luo, et al. Hybridna: A hybrid transformer-mamba2 long-range dna language model.arXiv preprint arXiv:2502.10807,

    work page arXiv

  26. [26]

    Jepa-dna: Grounding genomic foundation models through joint-embedding predictive architectures.arXiv preprint arXiv:2602.17162,

    Ariel Larey, Elay Dahan, Amit Bleiweiss, Raizy Kellerman, Guy Leib, Omri Nayshool, Dan Ofer, Tal Zinger, Dan Dominissini, Gideon Rechavi, et al. Jepa-dna: Grounding genomic foundation models through joint-embedding predictive architectures.arXiv preprint arXiv:2602.17162,

    work page arXiv

  27. [27]

    Generator: a long-context generative genomic foundation model

    13 Wei Wu, Qiuyi Li, Yuanyuan Zhang, Zhihao Zhan, Ruipu Chen, Mingyang Li, Kun Fu, Junyan Qi, Yongzhou Bao, Chao Wang, et al. Generator: a long-context generative genomic foundation model. arXiv preprint arXiv:2502.07272,

    work page arXiv

  28. [28]

    Towards on-device foundation models for raw wearable signals

    Simon A Lee, Cyrus Tanade, Hao Zhou, Juhyeon Lee, Megha Thukral, Baiying Lu, and Sharanya Ar- cot Desai. Towards on-device foundation models for raw wearable signals. InNeurIPS 2025 Workshop on Learning from Time Series for Health,

    work page 2025

  29. [29]

    Raptor: Scalable train-free embeddings for 3d medical volumes leveraging pretrained 2d foundation models.arXiv preprint arXiv:2507.08254,

    Ulzee An, Moonseong Jeong, Simon A Lee, Aditya Gorla, Yuzhe Yang, and Sriram Sankararaman. Raptor: Scalable train-free embeddings for 3d medical volumes leveraging pretrained 2d foundation models.arXiv preprint arXiv:2507.08254,

    work page arXiv

  30. [30]

    E3d-gpt: Enhanced 3d visual foundation for medical vision-language model.arXiv preprint arXiv:2410.14200,

    Haoran Lai, Zihang Jiang, Qingsong Yao, Rongsheng Wang, Zhiyang He, Xiaodong Tao, Wei Wei, Weifu Lv, and S Kevin Zhou. E3d-gpt: Enhanced 3d visual foundation for medical vision-language model.arXiv preprint arXiv:2410.14200,

    work page arXiv

  31. [31]

    Zixuan Liu, Hanwen Xu, Addie Woicik, Linda G Shapiro, Marian Blazes, Yue Wu, Cecilia S Lee, Aaron Y Lee, and Sheng Wang. Octcube: a 3d foundation model for optical coherence tomography that improves cross-dataset, cross-disease, cross-device and cross-modality analysis.arXiv preprint arXiv:2408.11227, 2024b. Abdelrahman M Shaker, Muhammad Maaz, Hanoona Ra...

    work page arXiv

  32. [32]

    Clinical modernbert: An efficient and long context encoder for biomedical text.arXiv preprint arXiv:2504.03964, April

    Simon A Lee, Anthony Wu, and Jeffrey N Chiang. Clinical modernbert: An efficient and long context encoder for biomedical text.arXiv preprint arXiv:2504.03964, April

    work page arXiv

  33. [33]

    A Survey of Large Language Models

    Wayne Xin Zhao, Kun Zhou, Junyi Li, Tianyi Tang, Xiaolei Wang, Yupeng Hou, Yingqian Min, Beichen Zhang, Junjie Zhang, Zican Dong, et al. A survey of large language models.arXiv preprint arXiv:2303.18223, 1(2),

    work page internal anchor Pith review Pith/arXiv arXiv

  34. [34]

    Medical event data standard (meds): Facilitating machine learning for health

    14 Bert Arnrich, Edward Choi, Jason Alan Fries, Matthew BA McDermott, Jungwoo Oh, Tom Pollard, Nigam Shah, Ethan Steinberg, Michael Wornow, and Robin van de Water. Medical event data standard (meds): Facilitating machine learning for health. InICLR 2024 Workshop on Learning from Time Series For Health, pages 03–08,

    work page 2024

  35. [35]

    Medhelm: Holistic evaluation of large language models for medical tasks.arXiv preprint arXiv:2505.23802,

    Suhana Bedi, Hejie Cui, Miguel Fuentes, Alyssa Unell, Michael Wornow, Juan M Banda, Nikesh Kotecha, Timothy Keyes, Yifan Mai, Mert Oez, et al. Medhelm: Holistic evaluation of large language models for medical tasks.arXiv preprint arXiv:2505.23802,

    work page arXiv

  36. [36]

    HealthBench: Evaluating Large Language Models Towards Improved Human Health

    Rahul K Arora, Jason Wei, Rebecca Soskin Hicks, Preston Bowman, Joaquin Quiñonero-Candela, Foivos Tsimpourlas, Michael Sharman, Meghan Shah, Andrea Vallone, Alex Beutel, et al. Health- bench: Evaluating large language models towards improved human health.arXiv preprint arXiv:2505.08775,

    work page internal anchor Pith review Pith/arXiv arXiv