Pith. sign in

REVIEW 3 major objections 2 minor 109 references

Spatial navigation tasks can flag Alzheimer's risk years before memory symptoms, tracking early pathology biomarkers in people who still test as cognitively normal.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · grok-4.5

2026-07-13 19:52 UTC pith:VYMPWEAT

load-bearing objection Useful narrative synthesis on navigation as a preclinical AD marker, but I only have the abstract—and the attached full text is the wrong paper—so treat claims about p-tau correlations and screening utility as still unchecked. the 3 major comments →

arxiv 2603.23082 v2 pith:VYMPWEAT submitted 2026-03-24 q-bio.NC

Spatial navigation in preclinical Alzheimer's disease: A review

Syrine Salouhou , Victor Gilles , Remi Vall\'ee , Gillian T. Coughlan , Romain Bachelet , Michael Hornberger , Hugo Spiers , Antoine Coutrot
show 1 more author
Antoine Garnier-Crussard
This is my paper
classification q-bio.NC
keywords preclinical Alzheimer's diseasespatial navigationpath integrationwayfindingp-taubiomarkersearly detectionmedial temporal lobe
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

Alzheimer's disease builds quietly for years before memory tests fail. This review argues that spatial navigation—especially path integration and wayfinding—is hit earlier than episodic memory because it depends on the same brain circuits where AD pathology first appears. Drawing on studies of cognitively unimpaired people who already carry AD biomarkers, it shows that navigation performance lines up with plasma and CSF markers, notably p-tau. If that link holds, simple, scalable navigation tests could identify people at risk in the preclinical window and guide earlier interventions before clinically significant decline sets in.

Core claim

In cognitively unimpaired individuals with AD biomarkers, performance on spatial navigation tasks—particularly path integration and wayfinding—correlates with plasma and CSF markers of AD pathology (notably p-tau), making navigation assessment a sensitive candidate for preclinical risk detection.

What carries the argument

The alignment of spatial navigation computations (path integration and wayfinding) with the neural circuits that are the earliest sites of AD pathology—used to explain why navigation measures can detect risk before episodic memory decline.

Load-bearing premise

That the observed links between navigation scores and AD biomarkers in still-unimpaired people truly reflect early AD circuit damage, not age, vascular disease, education, task quirks, or other pathology—and that those links will predict future clinical decline.

What would settle it

A prospective study in cognitively unimpaired biomarker-positive people showing that baseline path-integration or wayfinding scores do not predict later cognitive decline or conversion once age, vascular burden, education, and non-AD pathology are controlled.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Path integration and wayfinding tests could be added to preclinical screening batteries alongside or ahead of standard memory tests.
  • Plasma or CSF p-tau levels may be interpretable together with navigation scores as a joint early-risk signal.
  • Scalable navigation assessments could help select asymptomatic at-risk participants for prevention trials.
  • Interventions timed to navigation decline might slow progression before clinically significant impairment appears.
  • Future work should prioritize longitudinal designs that test whether navigation change predicts clinical outcomes.

Where Pith is reading between the lines

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

  • If navigation truly tracks earliest circuit failure, digital or VR path-integration tasks could become remote, low-cost screening tools outside specialty clinics.
  • Dissociating path integration from wayfinding may help separate entorhinal-centered from broader network contributions to early AD risk.
  • Combining navigation metrics with plasma p-tau could tighten enrichment of prevention trials beyond biomarkers alone.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 2 minor

Summary. This manuscript is submitted as a narrative review arguing that spatial navigation—especially path integration and wayfinding—is a particularly sensitive cognitive marker in preclinical Alzheimer’s disease. The abstract claims that navigation depends on circuits that are among the earliest sites of AD pathology, that performance on such tasks correlates with plasma and CSF AD biomarkers (notably p-tau) in cognitively unimpaired biomarker-positive individuals, and that navigation assessment could therefore provide a sensitive, scalable approach for early risk detection and for informing future interventions. The abstract further contrasts this putative early sensitivity with episodic memory decline, which is said to appear only after more substantial medial temporal damage.

Significance. If the synthesis is accurate and the cited associations are robust after appropriate confound control and prospective validation, the review would be clinically and scientifically useful: it would consolidate a circuit-to-cognition rationale for navigation as a preclinical marker, highlight task classes (path integration, wayfinding) that may outperform standard memory tests at the asymptomatic stage, and motivate scalable digital or behavioral screening. That contribution would be of clear interest to cognitive neuroscience and AD biomarker communities. However, the significance of the present submission cannot be assessed from the materials provided, because the body text supplied under this arXiv ID does not correspond to the titled review.

major comments (3)
  1. The full manuscript text attached under paper_id 2603.23082 / title “Spatial navigation in preclinical Alzheimer’s disease: A review” is not that review. The body is an unrelated methods paper on MedCausalX (adaptive causal reasoning for medical vision–language models; arXiv 2603.23085), with its own abstract, CRMed dataset, reflective tokens, DPO/GRPO pipeline, and medical VQA tables. None of the promised content—overview of spatial navigation computations and tasks, mapping to earliest AD pathology sites, or synthesis of cognitively unimpaired biomarker-positive cohorts—is present. Peer review of the stated central claim is therefore impossible on the supplied file.
  2. Even taking only the abstract of the intended review as the claim set, the load-bearing assertion that path integration and wayfinding performance “correlates with plasma and CSF biomarkers of AD pathology, notably p-tau” in cognitively unimpaired at-risk individuals, and that this “can represent a sensitive and scalable approach for early detection,” cannot be checked: there is no methods section defining inclusion criteria, no table of primary studies with effect sizes or quality appraisal, no discussion of confounds (age, vascular burden, education, task design, non-AD pathology), and no longitudinal outcome data supporting prospective risk prediction. Those elements are required for a review whose clinical recommendation rests on that correlation.
  3. The abstract’s mechanistic premise—that navigation relies on neural circuits corresponding to the earliest sites of AD pathology and is therefore more sensitive than episodic memory—is a central organizing claim. Without the actual review body (circuit mapping, staging of pathology, and head-to-head comparison with memory measures in the same biomarker-defined cohorts), this premise remains an unexamined assertion rather than a documented synthesis.
minor comments (2)
  1. Abstract wording is generally clear, but “individuals at-risk of AD” should be defined more precisely (e.g., amyloid/tau biomarker criteria, genetic risk, or both) once the correct full text is available.
  2. The abstract asserts clinical utility (“will inform future interventions”) without qualifying the current evidence level (cross-sectional association vs. prospective prediction). Softening that language until longitudinal data are reviewed would improve accuracy.

Circularity Check

0 steps flagged

Narrative review of external navigation–biomarker correlations; no derivation-by-construction or load-bearing circular step.

full rationale

Paper 2603.23082 is a narrative review whose strongest claim is that path-integration and wayfinding performance correlates with plasma/CSF AD biomarkers (notably p-tau) in cognitively unimpaired at-risk individuals, and may aid preclinical detection. That claim is framed as a synthesis of external empirical studies, not as a first-principles derivation, fitted parameter renamed as prediction, or uniqueness theorem imported from the authors. There is no equation chain in which navigation scores are defined from the same biomarkers they are said to predict, nor any self-definitional loop (X defined via Y then used to derive Y). Residual risks for a review of this type—selective citation, cross-sectional confounds, lack of new longitudinal outcome data—are correctness/generalization concerns, not circularity under the enumerated patterns. The supplied full-text block is an unrelated MedCausalX VLM manuscript and cannot be used to invent circular steps for 2603.23082. Honest finding: no significant circularity; score 0; steps empty.

Axiom & Free-Parameter Ledger

0 free parameters · 4 axioms · 0 invented entities

As a clinical neuroscience review (abstract only), load-bearing premises are domain assumptions about AD staging, circuit mapping, and biomarker validity rather than free parameters or invented physical entities. No fitted constants appear in the abstract. Invented entities are absent; the paper reuses established constructs (path integration, wayfinding, p-tau, preclinical AD).

axioms (4)
  • domain assumption AD has a prolonged preclinical phase in which neuropathology accumulates before clinical cognitive symptoms.
    Stated in the abstract as the premise for seeking early cognitive markers; standard in modern AD research but still a staging model with evolving biomarker criteria.
  • domain assumption Spatial navigation depends on neural circuits that are among the earliest sites of AD pathology (e.g., medial temporal / entorhinal-related systems).
    Core mechanistic link asserted in the abstract; strength depends on how tightly the review maps specific tasks to specific pathology sites.
  • domain assumption Plasma and CSF p-tau (and related AD biomarkers) validly index AD pathology in cognitively unimpaired individuals.
    The claimed navigation–biomarker correlations inherit the validity and specificity limits of those biomarkers.
  • domain assumption Episodic memory decline typically appears only after substantial medial temporal damage, making it less sensitive preclinically than navigation.
    Comparative sensitivity claim used to motivate navigation as an earlier marker; requires careful task- and sample-matched evidence.

pith-pipeline@v1.1.0-grok45 · 29667 in / 2531 out tokens · 27912 ms · 2026-07-13T19:52:31.230461+00:00 · methodology

0 comments
read the original abstract

Alzheimer's disease (AD) develops over a prolonged preclinical phase, during which neuropathological changes accumulate long before cognitive symptoms appear. Identifying cognitive functions affected at early stages is critical for the preclinical detection of asymptomatic individuals at-risk of AD. Early risk identification could enable timely interventions aimed at mitigating the development of significant future cognitive impairment. While episodic memory decline typically appears after substantial medial temporal lobe damage, spatial navigation has emerged as a particularly sensitive cognitive function in preclinical AD. In this review, we provide an overview of spatial navigation computations and the tasks used to assess them, highlighting how spatial navigation relies on neural circuits corresponding to the earliest sites of AD pathology. We synthesize evidence from cognitively unimpaired individuals with AD biomarkers, i.e. individuals at-risk of AD, and discuss future research directions. Overall, performance on spatial navigation tasks, particularly path integration and wayfinding, correlates with plasma and CSF biomarkers of AD pathology, notably p-tau. Spatial navigation assessment can represent a sensitive and scalable approach for early detection of individuals at-risk of AD in preclinical stages, and will inform future interventions to mitigate the progression toward clinically significant cognitive impairment.

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

109 extracted references · 31 linked inside Pith

  1. [1]

    Gpt-4 technical report.arXiv preprint arXiv:2303.08774, 2023

    Josh Achiam, Steven Adler, Sandhini Agarwal, Lama Ahmad, Ilge Akkaya, Florencia Leoni Aleman, Diogo Almeida, Janko Altenschmidt, Sam Altman, Shyamal Anadkat, et al. Gpt-4 technical report.arXiv preprint arXiv:2303.08774, 2023. 1

    Pith/arXiv arXiv 2023

  2. [2]

    Qwen2.5-vl: Technical report

    Shuai Bai and et al. Qwen2.5-vl: Technical report. https : / / arxiv . org / abs / 2502 . 13923, 2025. arXiv preprint arXiv:2502.13923. 5, 6, 7, 9

    Pith/arXiv arXiv 2025

  3. [3]

    Constitutional ai: Harmlessness from ai feedback, 2022

    Yuntao Bai, Saurav Kadavath, Sandipan Kundu, Amanda Askell, Jackson Kernion, Andy Jones, Anna Chen, Anna Goldie, Azalia Mirhoseini, Cameron McKinnon, and et al. Constitutional ai: Harmlessness from ai feedback, 2022. 3

    2022

  4. [4]

    Coun- terfactual causal-effect intervention for interpretable medi- cal visual question answering.IEEE Transactions on Med- ical Imaging, 2024

    Linqin Cai, Haodu Fang, Nuoying Xu, and Bo Ren. Coun- terfactual causal-effect intervention for interpretable medi- cal visual question answering.IEEE Transactions on Med- ical Imaging, 2024. 3

    2024

  5. [5]

    Causality matters in medical imaging.Nature Communications, 11 (1):3673, 2020

    Daniel C Castro, Ian Walker, and Ben Glocker. Causality matters in medical imaging.Nature Communications, 11 (1):3673, 2020. 2

    2020

  6. [6]

    360+x: A panoptic multi- modal scene understanding dataset

    Hao Chen, Yuqi Hou, Chenyuan Qu, Irene Testini, Xiao- han Hong, and Jianbo Jiao. 360+x: A panoptic multi- modal scene understanding dataset. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 19373–19382, 2024. 3

    2024

  7. [7]

    Huatuogpt-vision, to- wards injecting medical visual knowledge into multimodal llms at scale.arXiv preprint arXiv:2406.19280, 2024

    Junying Chen, Chi Gui, Ruyi Ouyang, Anningzhe Gao, Shunian Chen, Guiming Hardy Chen, Xidong Wang, Ruifei Zhang, Zhenyang Cai, Ke Ji, et al. Huatuogpt-vision, to- wards injecting medical visual knowledge into multimodal llms at scale.arXiv preprint arXiv:2406.19280, 2024. 3

    Pith/arXiv arXiv 2024

  8. [8]

    Teaching large language models to self-debug

    Xinyun Chen, Maxwell Lin, Nathanael Sch ¨arli, and Denny Zhou. Teaching large language models to self-debug. In ICLR, 2024. 2

    2024

  9. [9]

    Bridging radiology and pathology foundation models via concept- based multimodal co-adaptation

    Yihang Chen, Yanyan Huang, Fuying Wang, Maximus Ye- ung, Yuming Jiang, Shujun Wang, and Lequan Yu. Bridging radiology and pathology foundation models via concept- based multimodal co-adaptation. InThe Fourteenth Inter- national Conference on Learning Representations. 1

  10. [10]

    Expanding performance boundaries of open-source multimodal models with model, data, and test- time scaling.arXiv preprint arXiv:2412.05271, 2024

    Zhe Chen, Weiyun Wang, Yue Cao, Yangzhou Liu, Zhang- wei Gao, Erfei Cui, Jinguo Zhu, Shenglong Ye, Hao Tian, Zhaoyang Liu, et al. Expanding performance boundaries of open-source multimodal models with model, data, and test- time scaling.arXiv preprint arXiv:2412.05271, 2024. 5, 6, 11, 12

    Pith/arXiv arXiv 2024

  11. [11]

    Internvl: Scaling up vision founda- tion models and aligning for generic visual-linguistic tasks

    Zhe Chen, Jiannan Wu, Wenhai Wang, Weijie Su, Guo Chen, Sen Xing, Muyan Zhong, Qinglong Zhang, Xizhou Zhu, Lewei Lu, et al. Internvl: Scaling up vision founda- tion models and aligning for generic visual-linguistic tasks. InProceedings of the IEEE/CVF conference on computer vision and pattern recognition, pages 24185–24198, 2024. 7

    2024

  12. [12]

    Deep reinforcement learn- ing from human preferences.Advances in neural informa- tion processing systems, 30, 2017

    Paul F Christiano, Jan Leike, Tom Brown, Miljan Martic, Shane Legg, and Dario Amodei. Deep reinforcement learn- ing from human preferences.Advances in neural informa- tion processing systems, 30, 2017. 3

    2017

  13. [13]

    Tinymig: Trans- ferring generalization from vision foundation models to single-domain medical imaging

    LIU Chuang, Hongyan Xu, Yichao Cao, Xiu Su, Zhe Qu, Tianfa Li, Shan An, and Haogang Zhu. Tinymig: Trans- ferring generalization from vision foundation models to single-domain medical imaging. InForty-second Interna- tional Conference on Machine Learning, 2025. 1

    2025

  14. [14]

    Gaussiandwm: 3d gaussian driving world model for unified scene understanding and multi-modal generation.arXiv preprint arXiv:2512.23180,

    Tianchen Deng, Xuefeng Chen, Yi Chen, Qu Chen, Yuyao Xu, Lijin Yang, Le Xu, Yu Zhang, Bo Zhang, Wuxiong Huang, and Hesheng Wang. Gaussiandwm: 3d gaussian driving world model for unified scene understanding and multi-modal generation.arXiv preprint arXiv:2512.23180,

    Pith/arXiv arXiv

  15. [15]

    What is the best 3d scene representation for robotics? from geometric to foundation models.arXiv preprint arXiv:2512.03422, 2025

    Tianchen Deng, Yue Pan, Shenghai Yuan, Dong Li, Chen Wang, Mingrui Li, Long Chen, Lihua Xie, Danwei Wang, Jingchuan Wang, Javier Civera, Hesheng Wang, and Wei- dong Chen. What is the best 3d scene representation for robotics? from geometric to foundation models.arXiv preprint arXiv:2512.03422, 2025. 3

    arXiv 2025

  16. [16]

    En- hancing large vision language models with self-training on image comprehension.Advances in Neural Information Processing Systems, 37:131369–131397, 2024

    Yihe Deng, Pan Lu, Fan Yin, Ziniu Hu, Sheng Shen, Quan- quan Gu, James Zou, Kai-Wei Chang, and Wei Wang. En- hancing large vision language models with self-training on image comprehension.Advances in Neural Information Processing Systems, 37:131369–131397, 2024. 1

    2024

  17. [17]

    Prpo: Aligning process reward with outcome reward in policy optimization

    Ruiyi Ding, Yongxuan Lv, Xianhui Meng, Jiahe Song, Chao Wang, Chen Jiang, and Yuan Cheng. Prpo: Aligning process reward with outcome reward in policy optimization. arXiv preprint arXiv:2601.07182, 2026. 3

    arXiv 2026

  18. [18]

    Neurosymb-mrg: Differentiable abductive reasoning with active uncertainty minimization for radiology report generation.arXiv preprint arXiv:2603.01756, 2026

    Rong Fu, Yiqing Lyu, Chunlei Meng, Muge Qi, Yabin Jin, Qi Zhao, Li Bao, Juntao Gao, Fuqian Shi, Nilanjan Dey, et al. Neurosymb-mrg: Differentiable abductive reasoning with active uncertainty minimization for radiology report generation.arXiv preprint arXiv:2603.01756, 2026. 3

    Pith/arXiv arXiv 2026

  19. [19]

    Shortcut learning in deep neural net- works.Nature Machine Intelligence, 2(11):665–673, 2020

    Robert Geirhos, J ¨orn-Henrik Jacobsen, Claudio Michaelis, Richard Zemel, Wieland Brendel, Matthias Bethge, and Fe- lix A Wichmann. Shortcut learning in deep neural net- works.Nature Machine Intelligence, 2(11):665–673, 2020. 1, 2

    2020

  20. [20]

    Med-cmr: A fine-grained benchmark integrating visual evidence and clinical logic for medical complex multimodal reasoning.arXiv preprint arXiv:2512.00818, 2025

    Haozhen Gong, Xiaozhong Ji, Yuansen Liu, Wenbin Wu, Xiaoxiao Yan, Jingjing Liu, Kai Wu, Jiazhen Pan, Bailiang Jian, Jiangning Zhang, et al. Med-cmr: A fine-grained benchmark integrating visual evidence and clinical logic for medical complex multimodal reasoning.arXiv preprint arXiv:2512.00818, 2025. 1

    arXiv 2025

  21. [21]

    Deepseek-r1: Incentivizing reason- ing capability in llms via reinforcement learning.arXiv preprint arXiv:2501.12948, 2025

    Daya Guo, Dejian Yang, Haowei Zhang, Junxiao Song, Ruoyu Zhang, Runxin Xu, Qihao Zhu, Shirong Ma, Peiyi Wang, Xiao Bi, et al. Deepseek-r1: Incentivizing reason- ing capability in llms via reinforcement learning.arXiv preprint arXiv:2501.12948, 2025. 3

    Pith/arXiv arXiv 2025

  22. [22]

    Meddr: Diagnosis-guided bootstrapping for large-scale medical vision-language learning.arXiv preprint arXiv:2404.15127, 2024

    Sunan He, Yuxiang Nie, Zhixuan Chen, Zhiyuan Cai, Hongmei Wang, Shu Yang, and Hao Chen. Meddr: Diagnosis-guided bootstrapping for large-scale medical vision-language learning.arXiv preprint arXiv:2404.15127, 2024. 5, 6, 7

    Pith/arXiv arXiv 2024

  23. [23]

    Pathological visual ques- tion answering.arXiv preprint arXiv:2010.12435, 2020

    Xuehai He, Zhuo Cai, Wenlan Wei, Yichen Zhang, Luntian Mou, Eric Xing, and Pengtao Xie. Pathological visual ques- tion answering.arXiv preprint arXiv:2010.12435, 2020. 5, 6, 9

    Pith/arXiv arXiv 2010

  24. [24]

    Codev: Code with images for faithful visual reasoning via tool-aware policy optimization.arXiv preprint arXiv:2511.19661, 2025

    Xinhai Hou, Shaoyuan Xu, Manan Biyani, Moyan Li, Jia Liu, Todd C Hollon, and Bryan Wang. Codev: Code with images for faithful visual reasoning via tool-aware policy optimization.arXiv preprint arXiv:2511.19661, 2025. 2, 3

    arXiv 2025

  25. [25]

    Advancing medical imaging with language mod- els: A journey from n-grams to chatgpt.arXiv preprint arXiv:2304.04920, 2023

    Mingzhe Hu, Shaoyan Pan, Yuheng Li, and Xiaofeng Yang. Advancing medical imaging with language mod- els: A journey from n-grams to chatgpt.arXiv preprint arXiv:2304.04920, 2023. 3

    Pith/arXiv arXiv 2023

  26. [26]

    Large language models cannot self-correct reasoning yet

    Jie Huang, Shixiang Shane Gu, Le Hou, Yuexin Wu, Xuezhi Wang, Hongkun Yu, and Jiawei Han. Large language models cannot self-correct reasoning yet. In Proceedings of the International Conference on Learn- ing Representations (ICLR) 2024, 2024. Paper ID: 8b4add8b0aa8749d80a34ca5d941c355. 2

    2024

  27. [27]

    Medreflect: Teaching medical llms to self-improve via reflective correc- tion.arXiv preprint arXiv:2510.03687, 2025

    Yue Huang, Yanyuan Chen, Dexuan Xu, Weihua Yue, Huamin Zhang, Meikang Qiu, and Yu Huang. Medreflect: Teaching medical llms to self-improve via reflective correc- tion.arXiv preprint arXiv:2510.03687, 2025. 1, 2

    arXiv 2025

  28. [28]

    Gpt-4o system card.arXiv preprint arXiv:2410.21276, 2024

    Aaron Hurst, Adam Lerer, Adam P Goucher, Adam Perel- man, Aditya Ramesh, Aidan Clark, AJ Ostrow, Akila Weli- hinda, Alan Hayes, Alec Radford, et al. Gpt-4o system card.arXiv preprint arXiv:2410.21276, 2024. 5, 12

    Pith/arXiv arXiv 2024

  29. [29]

    Med-moe: Mixture of domain-specific experts for lightweight medical vision-language models

    Songtao Jiang, Tuo Zheng, Yan Zhang, Yeying Jin, Li Yuan, and Zuozhu Liu. Med-moe: Mixture of domain-specific experts for lightweight medical vision-language models. In Findings of the association for computational linguistics: EMNLP 2024, pages 3843–3860, 2024. 3

    2024

  30. [30]

    Hulu-med: A transparent general- ist model towards holistic medical vision-language under- standing.arXiv preprint arXiv:2510.08668, 2025

    Songtao Jiang, Yuan Wang, Sibo Song, Tianxiang Hu, Chenyi Zhou, Bin Pu, Yan Zhang, Zhibo Yang, Yang Feng, Joey Tianyi Zhou, et al. Hulu-med: A transparent general- ist model towards holistic medical vision-language under- standing.arXiv preprint arXiv:2510.08668, 2025. 3

    arXiv 2025

  31. [31]

    Why adam can beat sgd: Second-moment normalization yields sharper tails.arXiv preprint arXiv:2603.03099, 2026

    Ruinan Jin, Yingbin Liang, and Shaofeng Zou. Why adam can beat sgd: Second-moment normalization yields sharper tails.arXiv preprint arXiv:2603.03099, 2026. 6

    Pith/arXiv arXiv 2026

  32. [32]

    Tic-grpo: Provable and efficient optimiza- tion for reinforcement learning from human feedback

    Ruinan Jin et al. Tic-grpo: Provable and efficient optimiza- tion for reinforcement learning from human feedback. 3

  33. [33]

    Mimic-cxr, a de- identified publicly available database of chest radiographs with free-text reports.Scientific data, 6(1):317, 2019

    Alistair EW Johnson, Tom J Pollard, Seth J Berkowitz, Nathaniel R Greenbaum, Matthew P Lungren, Chih-ying Deng, Roger G Mark, and Steven Horng. Mimic-cxr, a de- identified publicly available database of chest radiographs with free-text reports.Scientific data, 6(1):317, 2019. 5, 9

    2019

  34. [34]

    Large language models are zero-shot reasoners.Advances in Neural Information Pro- cessing Systems, 35:22199–22213, 2022

    Takeshi Kojima, Shixiang Shane Gu, Machel Reid, Yutaka Matsuo, and Yusuke Iwasawa. Large language models are zero-shot reasoners.Advances in Neural Information Pro- cessing Systems, 35:22199–22213, 2022. 2

    2022

  35. [35]

    Co-Reyes, Avi Singh, Kate Baumli, Shariq Iqbal, Colton Bishop, Rebecca Roelofs, Lei M

    Aviral Kumar, Vincent Zhuang, Rishabh Agarwal, Yi Su, John D. Co-Reyes, Avi Singh, Kate Baumli, Shariq Iqbal, Colton Bishop, Rebecca Roelofs, Lei M. Zhang, Kay McK- inney, Disha Shrivastava, Cosmin Paduraru, George Tucker, Doina Precup, Feryal Behbahani, and Aleksandra Faust. Training language models to self-correct via reinforcement learning. 2024. 2

    2024

  36. [36]

    Med-r1: Re- inforcement learning for generalizable medical reasoning in vision-language models.IEEE Transactions on Medical Imaging, 2026

    Yuxiang Lai, Jike Zhong, Ming Li, Shitian Zhao, Yuheng Li, Konstantinos Psounis, and Xiaofeng Yang. Med-r1: Re- inforcement learning for generalizable medical reasoning in vision-language models.IEEE Transactions on Medical Imaging, 2026. 1, 2, 3, 5, 6, 11, 12

    2026

  37. [37]

    A dataset of clinically generated visual questions and answers about radiology images.Scientific data, 5(1):1–10, 2018

    Jason J Lau, Soumya Gayen, Asma Ben Abacha, and Dina Demner-Fushman. A dataset of clinically generated visual questions and answers about radiology images.Scientific data, 5(1):1–10, 2018. 5, 6, 9

    2018

  38. [38]

    Exploring efficient open- vocabulary segmentation in the remote sensing.arXiv preprint arXiv:2509.12040, 2025

    Bingyu Li, Haocheng Dong, Da Zhang, Zhiyuan Zhao, Junyu Gao, and Xuelong Li. Exploring efficient open- vocabulary segmentation in the remote sensing.arXiv preprint arXiv:2509.12040, 2025. 3

    arXiv 2025

  39. [39]

    Maris: Marine open-vocabulary in- stance segmentation with geometric enhancement and se- mantic alignment.arXiv preprint arXiv:2510.15398, 2025

    Bingyu Li, Feiyu Wang, Da Zhang, Zhiyuan Zhao, Junyu Gao, and Xuelong Li. Maris: Marine open-vocabulary in- stance segmentation with geometric enhancement and se- mantic alignment.arXiv preprint arXiv:2510.15398, 2025

    arXiv 2025

  40. [40]

    Stitchfusion: Weaving any visual modalities to en- hance multimodal semantic segmentation

    Bingyu Li, Da Zhang, Zhiyuan Zhao, Junyu Gao, and Xue- long Li. Stitchfusion: Weaving any visual modalities to en- hance multimodal semantic segmentation. InProceedings of the 33rd ACM International Conference on Multimedia, pages 1308–1317, 2025. 3

    2025

  41. [41]

    Llava-med: Training a large language-and-vision assistant for biomedicine in one day

    Chunyuan Li, Cliff Wong, Sheng Zhang, Naoto Usuyama, Haotian Liu, Jianwei Yang, Tristan Naumann, Hoifung Poon, and Jianfeng Gao. Llava-med: Training a large language-and-vision assistant for biomedicine in one day. Advances in Neural Information Processing Systems, 36: 28541–28564, 2023. 5, 6, 12

    2023

  42. [42]

    The choice of divergence: A neglected key to mitigating diversity collapse in reinforcement learning with verifiable reward.arXiv preprint arXiv:2509.07430,

    Long Li, Jiaran Hao, Jason Klein Liu, Zhijian Zhou, Yant- ing Miao, Wei Pang, Xiaoyu Tan, Wei Chu, Zhe Wang, Shirui Pan, et al. The choice of divergence: A neglected key to mitigating diversity collapse in reinforcement learning with verifiable reward.arXiv preprint arXiv:2509.07430,

    arXiv

  43. [43]

    M²IV: Towards efficient and fine-grained multimodal in-context learning via representation engineering

    Yanshu Li, Yi Cao, Hongyang He, Qisen Cheng, Xiang Fu, Xi Xiao, Tianyang Wang, and Ruixiang Tang. M²IV: Towards efficient and fine-grained multimodal in-context learning via representation engineering. InSecond Con- ference on Language Modeling, 2025. 3

    2025

  44. [44]

    Mmt-ard: Multimodal multi-teacher adversarial dis- tillation for robust vision-language models.arXiv preprint arXiv:2511.17448, 2025

    Yuqi Li, Junhao Dong, Chuanguang Yang, Shiping Wen, Piotr Koniusz, Tingwen Huang, Yingli Tian, and Yew-Soon Ong. Mmt-ard: Multimodal multi-teacher adversarial dis- tillation for robust vision-language models.arXiv preprint arXiv:2511.17448, 2025. 3

    arXiv 2025

  45. [45]

    Catp: Contextually adap- tive token pruning for efficient and enhanced multimodal in-context learning.arXiv preprint arXiv:2508.07871,

    Yanshu Li, Jianjiang Yang, Zhennan Shen, Ligong Han, Haoyan Xu, and Ruixiang Tang. Catp: Contextually adap- tive token pruning for efficient and enhanced multimodal in-context learning.arXiv preprint arXiv:2508.07871,

    arXiv

  46. [46]

    Taco: Enhancing multimodal in-context learning via task mapping-guided sequence con- figuration

    Yanshu Li, Jianjiang Yang, Tian Yun, Pinyuan Feng, Jinfa Huang, and Ruixiang Tang. Taco: Enhancing multimodal in-context learning via task mapping-guided sequence con- figuration. InProceedings of the 2025 Conference on Em- pirical Methods in Natural Language Processing, pages 736–763, 2025. 3

    2025

  47. [47]

    Efficient Medical Image Segmentation via Reinforcement Learning-Driven K-Space Sampling.IEEE Transactions on Emerging Topics in Com- putational Intelligence, 2025

    Yuqi Li, Hansheng Zeng, Fuyan Zhang, Chuanguang Yang, Yanli Li, and Weiping Ding. Efficient Medical Image Segmentation via Reinforcement Learning-Driven K-Space Sampling.IEEE Transactions on Emerging Topics in Com- putational Intelligence, 2025. 3

    2025

  48. [48]

    Manxi Lin, Nina Weng, Kamil Mikolaj, Zahra Bashir, Morten B. S. Svendsen, Martin G. Tolsgaard, Anders Ny- mark Christensen, and Aasa Feragen. Shortcut learning in medical image segmentation. InMedical Image Computing and Computer Assisted Intervention – MICCAI 2024, pages 623–633. Springer, 2024. 1

    2024

  49. [49]

    Surgical post-training: Cutting er- rors, keeping knowledge.arXiv preprint arXiv:2603.01683,

    Wenye Lin and Kai Han. Surgical post-training: Cutting er- rors, keeping knowledge.arXiv preprint arXiv:2603.01683,

    Pith/arXiv arXiv

  50. [50]

    Gamebot: Transparent assessment of llm reasoning in games

    Wenye Lin, Jonathan Roberts, Yunhan Yang, Samuel Al- banie, Zongqing Lu, and Kai Han. Gamebot: Transparent assessment of llm reasoning in games. InProceedings of the 63rd Annual Meeting of the Association for Computa- tional Linguistics (Volume 1: Long Papers), pages 7656– 7682, 2025. 2

    2025

  51. [51]

    Slake: A semantically-labeled knowledge- enhanced dataset for medical visual question answering

    Bo Liu, Li-Ming Zhan, Li Xu, Lin Ma, Yan Yang, and Xiao-Ming Wu. Slake: A semantically-labeled knowledge- enhanced dataset for medical visual question answering. In 2021 IEEE 18th International Symposium on Biomedical Imaging (ISBI), pages 1650–1654. IEEE, 2021. 5, 6, 9

    2021

  52. [52]

    Perturbating, tuning, and collaborating: Har- nessing vision foundation models for single domain gen- eralization on medical imaging

    Chuang Liu, Yichao Cao, YingYing Zhang, Xiu Su, and Haogang Zhu. Perturbating, tuning, and collaborating: Har- nessing vision foundation models for single domain gen- eralization on medical imaging. InProceedings of the AAAI Conference on Artificial Intelligence, pages 5370– 5378, 2025. 1

    2025

  53. [53]

    Cdrrm: Contrast-driven rubric gener- ation for reliable and interpretable reward modeling.arXiv preprint arXiv:2603.08035, 2026

    Dengcan Liu, Fengkai Yang, Xiaohan Wang, Shurui Yan, Jiajun Chai, Jiahao Li, Yikun Ban, Zhendong Mao, Wei Lin, and Guojun Yin. Cdrrm: Contrast-driven rubric gener- ation for reliable and interpretable reward modeling.arXiv preprint arXiv:2603.08035, 2026. 3

    arXiv 2026

  54. [54]

    On the intrinsic self-correction capability of llms: Uncertainty and latent concept.arXiv preprint arXiv:2406.02378, 2024

    Guangliang Liu, Haitao Mao, Bochuan Cao, Zhiyu Xue, Xitong Zhang, Rongrong Wang, Jiliang Tang, and Kris- ten Johnson. On the intrinsic self-correction capability of llms: Uncertainty and latent concept.arXiv preprint arXiv:2406.02378, 2024. 1, 2

    Pith/arXiv arXiv 2024

  55. [55]

    Improved baselines with visual instruc- tion tuning.arXiv preprint arXiv:2310.03744, 2024

    Haotian Liu et al. Improved baselines with visual instruc- tion tuning.arXiv preprint arXiv:2310.03744, 2024. 11

    Pith/arXiv arXiv 2024

  56. [56]

    Medcot: Medical chain of thought via hierarchical expert

    Jiaxiang Liu, Yuan Wang, Jiawei Du, Joey Tianyi Zhou, and Zuozhu Liu. Medcot: Medical chain of thought via hierarchical expert. InProceedings of the 2024 Confer- ence on Empirical Methods in Natural Language Process- ing (EMNLP 2024), pages 17371–17389, 2024. 1, 3, 6

    2024

  57. [57]

    Automated opti- mization modeling via a localizable error-driven perspec- tive.arXiv preprint arXiv:2602.11164, 2026

    Weiting Liu, Han Wu, Yufei Kuang, Xiongwei Han, Tao Zhong, Jianfeng Feng, and Wenlian Lu. Automated opti- mization modeling via a localizable error-driven perspec- tive.arXiv preprint arXiv:2602.11164, 2026. 3

    arXiv 2026

  58. [58]

    Scanext: Enhancing 3d medical image segmentation with dual attention network and depth-wise convolution.He- liyon, 10(5), 2024

    Yajun Liu, Zenghui Zhang, Jiang Yue, and Weiwei Guo. Scanext: Enhancing 3d medical image segmentation with dual attention network and depth-wise convolution.He- liyon, 10(5), 2024. 3

    2024

  59. [59]

    M 3 hl: Mutual mask mix with high-low level feature consistency for semi-supervised medical im- age segmentation

    Yajun Liu, Zenghui Zhang, Jiang Yue, Weiwei Guo, and Dongying Li. M 3 hl: Mutual mask mix with high-low level feature consistency for semi-supervised medical im- age segmentation. InInternational Conference on Medi- cal Image Computing and Computer-Assisted Intervention, pages 314–323. Springer, 2025. 3

    2025

  60. [60]

    Decoupled weight decay regularization.arXiv preprint arXiv:1711.05101, 2017

    Ilya Loshchilov and Frank Hutter. Decoupled weight decay regularization.arXiv preprint arXiv:1711.05101, 2017. 6

    Pith/arXiv arXiv 2017

  61. [61]

    Gastric-x: A multimodal multi-phase bench- mark dataset for advancing vision-language models in gas- tric cancer analysis, 2026

    Sheng Lu, Hao Chen, Rui Yin, Juyan Ba, Yu Zhang, and Yuanzhe Li. Gastric-x: A multimodal multi-phase bench- mark dataset for advancing vision-language models in gas- tric cancer analysis, 2026. 3

    2026

  62. [62]

    Thinking with blueprints: Assisting vision-language models in spatial reasoning via structured object representation, 2026

    Weijian Ma, Shizhao Sun, Tianyu Yu, Ruiyu Wang, Tat- Seng Chua, and Jiang Bian. Thinking with blueprints: Assisting vision-language models in spatial reasoning via structured object representation, 2026. 2

    2026

  63. [63]

    Self- refine: Iterative refinement with self-feedback

    Aman Madaan, Niket Tandon, Prakhar Gupta, Skyler Hal- linan, Luyu Gao, Sarah Wiegreffe, Uri Alon, Nouha Dziri, Shrimai Prabhumoye, Yiming Yang, Shashank Gupta, Bodhisattwa Prasad Majumder, Katherine Hermann, Sean Welleck, Amir Yazdanbakhsh, and Peter Clark. Self- refine: Iterative refinement with self-feedback. InAdvances in Neural Information Processing ...

    2023

  64. [64]

    Med-flamingo: a multimodal medical few-shot learner

    Michael Moor, Qian Huang, Shirley Wu, Michihiro Ya- sunaga, Yash Dalmia, Jure Leskovec, Cyril Zakka, Ed- uardo Pontes Reis, and Pranav Rajpurkar. Med-flamingo: a multimodal medical few-shot learner. InMachine Learn- ing for Health (ML4H), pages 353–367. PMLR, 2023. 5, 6, 7

    2023

  65. [65]

    Self-taught self-correction for small language mod- els.arXiv preprint arXiv:2503.08681, 2025

    Viktor Moskvoretskii, Chris Biemann, and Irina Nik- ishina. Self-taught self-correction for small language mod- els.arXiv preprint arXiv:2503.08681, 2025. 2

    Pith/arXiv arXiv 2025

  66. [66]

    Training lan- guage models to follow instructions with human feedback

    Long Ouyang, Jeffrey Wu, Xu Jiang, Diogo Almeida, Car- roll Wainwright, Pamela Mishkin, Chong Zhang, Sandhini Agarwal, Katarina Slama, Alex Ray, et al. Training lan- guage models to follow instructions with human feedback. Advances in neural information processing systems, 35: 27730–27744, 2022. 3

    2022

  67. [67]

    Medvlm-r1: Incentivizing medical reasoning capability of vision-language models (vlms) via reinforce- ment learning

    Jiazhen Pan, Che Liu, Junde Wu, Fenglin Liu, Jiayuan Zhu, Hongwei Bran Li, Chen Chen, Cheng Ouyang, and Daniel Rueckert. Medvlm-r1: Incentivizing medical reasoning capability of vision-language models (vlms) via reinforce- ment learning. InInternational Conference on Medical Im- age Computing and Computer-Assisted Intervention, pages 337–347. Springer, 20...

    2025

  68. [68]

    Toward understanding why adam converges faster than sgd for transformers.arXiv preprint arXiv:2306.00204, 2023

    Yan Pan and Yuanzhi Li. Toward understanding why adam converges faster than sgd for transformers.arXiv preprint arXiv:2306.00204, 2023. 6

    Pith/arXiv arXiv 2023

  69. [69]

    On the theory and practice of grpo: A trajectory-corrected approach with fast conver- gence.arXiv preprint arXiv:2508.02833, 2025

    Lei Pang and Ruinan Jin. On the theory and practice of grpo: A trajectory-corrected approach with fast conver- gence.arXiv preprint arXiv:2508.02833, 2025. 3

    arXiv 2025

  70. [70]

    Elements of causal inference: foundations and learning al- gorithms

    Jonas Peters, Dominik Janzing, and Bernhard Scholkopf. Elements of causal inference: foundations and learning al- gorithms. MIT press, 2017. 2

    2017

  71. [71]

    Causal inference and counterfac- tual prediction in machine learning for actionable health- care.Nature Machine Intelligence, 2(7):369–375, 2020

    Mattia Prosperi, Yi Guo, Matt Sperrin, James S Koopman, Jae S Min, Xing He, Shannan Rich, Mo Wang, Iain E Buchan, and Jiang Bian. Causal inference and counterfac- tual prediction in machine learning for actionable health- care.Nature Machine Intelligence, 2(7):369–375, 2020. 2

    2020

  72. [72]

    Direct preference optimization: Your language model is secretly a reward model.Advances in neural information processing systems, 36:53728–53741, 2023

    Rafael Rafailov, Archit Sharma, Eric Mitchell, Christo- pher D Manning, Stefano Ermon, and Chelsea Finn. Direct preference optimization: Your language model is secretly a reward model.Advances in neural information processing systems, 36:53728–53741, 2023. 3, 9

    2023

  73. [73]

    Multimodal generative ai for medical image in- terpretation.Nature, 639(8056):888–896, 2025

    Vishwanatha M Rao, Michael Hla, Michael Moor, Subathra Adithan, Stephen Kwak, Eric J Topol, and Pranav Ra- jpurkar. Multimodal generative ai for medical image in- terpretation.Nature, 639(8056):888–896, 2025. 1

    2025

  74. [74]

    Im- proving the accuracy of medical diagnosis with causal ma- chine learning.Nature communications, 11(1):3923, 2020

    Jonathan G Richens, Ciar ´an M Lee, and Saurabh Johri. Im- proving the accuracy of medical diagnosis with causal ma- chine learning.Nature communications, 11(1):3923, 2020. 2

    2020

  75. [75]

    Proximal policy optimization algo- rithms.arXiv preprint arXiv:1707.06347, 2017

    John Schulman, Filip Wolski, Prafulla Dhariwal, Alec Rad- ford, and Oleg Klimov. Proximal policy optimization algo- rithms.arXiv preprint arXiv:1707.06347, 2017. 3

    Pith/arXiv arXiv 2017

  76. [76]

    Deepseekmath: Pushing the limits of mathematical reasoning in open language models.arXiv preprint arXiv:2402.03300, 2024

    Zhihong Shao, Peiyi Wang, Qihao Zhu, Runxin Xu, Junx- iao Song, Xiao Bi, Haowei Zhang, Mingchuan Zhang, YK Li, Yang Wu, et al. Deepseekmath: Pushing the limits of mathematical reasoning in open language models.arXiv preprint arXiv:2402.03300, 2024. 2, 3, 9

    Pith/arXiv arXiv 2024

  77. [77]

    Reflexion: Language agents with verbal reinforcement learning.Advances in Neural In- formation Processing Systems, 36:8634–8652, 2023

    Noah Shinn, Federico Cassano, Ashwin Gopinath, Karthik Narasimhan, and Shunyu Yao. Reflexion: Language agents with verbal reinforcement learning.Advances in Neural In- formation Processing Systems, 36:8634–8652, 2023. 2

    2023

  78. [78]

    Hume: Introducing system-2 thinking in visual- language-action model.arXiv preprint arXiv:2505.21432,

    Haoming Song, Delin Qu, Yuanqi Yao, Qizhi Chen, Qi Lv, Yiwen Tang, Modi Shi, Guanghui Ren, Maoqing Yao, Bin Zhao, et al. Hume: Introducing system-2 thinking in visual- language-action model.arXiv preprint arXiv:2505.21432,

    Pith/arXiv arXiv

  79. [79]

    Bottom-up policy optimization: Your language model policy secretly contains internal policies.arXiv preprint arXiv:2512.19673, 2025

    Yuqiao Tan, Minzheng Wang, Shizhu He, Huanxuan Liao, Chengfeng Zhao, Qiunan Lu, Tian Liang, Jun Zhao, and Kang Liu. Bottom-up policy optimization: Your language model policy secretly contains internal policies.arXiv preprint arXiv:2512.19673, 2025. 3

    Pith/arXiv arXiv 2025

  80. [80]

    Collaboration between clinicians and vision–language models in radiology report generation.Nature Medicine, 31 (2):599–608, 2025

    Ryutaro Tanno, David GT Barrett, Andrew Sellergren, Sumedh Ghaisas, Sumanth Dathathri, Abigail See, Jo- hannes Welbl, Charles Lau, Tao Tu, Shekoofeh Azizi, et al. Collaboration between clinicians and vision–language models in radiology report generation.Nature Medicine, 31 (2):599–608, 2025. 1

    2025

Showing first 80 references.