pith. sign in

arxiv: 2606.29763 · v1 · pith:37HP3BXInew · submitted 2026-06-29 · 💻 cs.CV · cs.AI

TopoAgent: An Agentic Framework for Automated Topology Learning in Medical Imaging

Guangyu Meng , Pengfei Gu , Xueyang Li , Yiyu Shi , Erin Wolf Chambers , Danny Z. Chen This is my paper

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

classification 💻 cs.CV cs.AI
keywords topological data analysispersistent homologymedical image analysisagentic frameworktopological descriptorsfeature extractionautomated selectionlarge language model agent
0
0 comments X

The pith

TopoAgent is an LLM-based agent that automatically picks the best topological descriptor for a medical image dataset and outputs the matching feature vectors with no task-specific training.

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

The paper sets out to show that an LLM agent equipped with domain tools and accumulating memory can examine medical images, reason about their topological traits, and choose among fifteen known descriptors the one that fits best. Conventional work fixes one descriptor such as persistence images for every task, so the diversity of possible topological representations remains unused. If the agent succeeds, downstream classifiers receive tailored feature vectors that capture connected components, loops, and shape properties that pixel-based networks overlook. The framework runs a Perception-Reasoning-Action-Reflection cycle supported by twenty-one tools whose capabilities were distilled from tests on twenty-six datasets and six classifiers. The result is an automated pipeline that adapts topology learning to each new collection of images.

Core claim

TopoAgent is an LLM-based agentic framework that automates topology learning for medical image analysis. It operates through a Perception-Reasoning-Action-Reflection loop supported by twenty-one domain-specific tools and dual memory that accumulates experience across runs. Its skill set is distilled from systematic evaluation of fifteen topological descriptors across twenty-six datasets with six classifiers. TopoAgent analyzes input images and their topological characteristics, reasons about which topological descriptors best suit the input, and determines the optimal descriptor and its configuration, all without task-specific training.

What carries the argument

The Perception-Reasoning-Action-Reflection loop with twenty-one domain-specific tools and dual memory that stores and reuses prior experience.

If this is right

  • The same agent can generate ready-to-use topological feature vectors for any new medical image collection.
  • No human expert or retraining step is required to switch descriptors when the dataset changes.
  • The method inherits the performance range observed across the fifteen descriptors evaluated on twenty-six datasets.
  • Topological information that pixel-level networks miss becomes available to any downstream classifier through the agent's output vectors.

Where Pith is reading between the lines

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

  • The same agent structure could be tested on non-medical image collections to see whether the tool set generalizes beyond the medical domain.
  • If memory across runs continues to improve selections, repeated use on similar hospital data streams might yield progressively better descriptor choices.
  • Pairing the agent's output vectors with existing deep networks could be checked to measure additive gains in classification tasks.

Load-bearing premise

An LLM agent supported only by domain tools and accumulated memory can reliably analyze images and select the optimal topological descriptor without any task-specific training.

What would settle it

Run the agent on a held-out medical image dataset, compare the downstream accuracy of its chosen descriptor against the accuracy obtained by always using the single most common fixed descriptor, and check whether the agent's selection is consistently no better or worse.

Figures

Figures reproduced from arXiv: 2606.29763 by Danny Z. Chen, Erin Wolf Chambers, Guangyu Meng, Pengfei Gu, Xueyang Li, Yiyu Shi.

Figure 1
Figure 1. Figure 1: (a) Balanced accuracy (%) of 6 descriptors (covering the top-3 per object type) on 5 datasets. PS = persistence statistics, PI = persistence image, TF = template func￾tion, ATOL = automatic topologically-oriented learning, MK = Minkowski functional, LBP = local binary pattern. The best descriptor (bold) differs across 5 datasets, con￾firming no single descriptor is best for all. (b) Reasoning comparison on… view at source ↗
Figure 2
Figure 2. Figure 2: An overview of the TopoAgent framework with four phases: Perception iden￾tifies the object type and analyzes persistent homology statistics; Reasoning proposes and determines the descriptor and parameters; Action calls the descriptor tools; Re￾flection validates the feature vector. The skill set S feeds descriptor properties and tiered rankings with parameters to Reasoning (dashed green). Short-term memory… view at source ↗
Figure 3
Figure 3. Figure 3: Examples of the skill set S using persistence image. Sprop encodes qualitative descriptor knowledge, Srank provides tiered rankings and reasoning chains, and Sparam specifies validated parameters per object type. 3.4 TopoBenchmark TopoBenchmark serves two purposes: it provides an empirical basis from which the skill set S is distilled, and it defines a frozen testbed for evaluating the agent’s descriptor d… view at source ↗
Figure 4
Figure 4. Figure 4: Two case studies on organ shapes with opposite PH profiles. Case A: sparse but meaningful PH (avg persistence=0.06) — the agent retains a Tier 4 proposal with data-driven parameters. Case B: noisy PH (avg persistence<0.01) — the agent abandons PH-based descriptors entirely. Same object type, same ranking, opposite determinations driven by image-specific PH signals. agent arrives at different determinations… view at source ↗
read the original abstract

Topological data analysis (TDA), particularly persistent homology (PH), captures geometric structural properties in medical images (e.g., connected components, loops, shape characteristics), which conventional pixel-level deep learning approaches often neglect. While many topological descriptors are known for converting persistence diagrams (PDs) or raw images into topological feature vectors, existing methods mostly default to a single fixed descriptor (e.g., persistence images), leaving the diversity of topological representations largely unexplored. To the best of our knowledge, there is no known large language model (LLM)-based agentic framework that can automatically determine the most suitable topological descriptors for a given image dataset and produce the corresponding topological feature vectors for downstream tasks. To fill this gap, we propose \textbf{TopoAgent}, an LLM-based agentic framework that automates topology learning for medical image analysis.TopoAgent operates through a Perception--Reasoning--Action--Reflection loop supported by 21 domain-specific tools and dual memory that accumulates experience across runs. Its skill set is distilled from systematic evaluation of 15 topological descriptors across 26 datasets with six classifiers. TopoAgent analyzes input images and their topological characteristics, reasons about which topological descriptors best suit the input, and determines the optimal descriptor and its configuration, all without task-specific training.

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 / 0 minor

Summary. The manuscript proposes TopoAgent, an LLM-based agentic framework for automating topology learning in medical imaging. It operates via a Perception–Reasoning–Action–Reflection loop with 21 domain-specific tools and dual memory, distilling skills from a systematic evaluation of 15 topological descriptors across 26 datasets with six classifiers. The framework is claimed to analyze input images, reason about suitable descriptors, select and configure the optimal one, and produce feature vectors for downstream tasks, all without task-specific training or human intervention.

Significance. If the central claim holds, the work would advance TDA applications in medical imaging by replacing fixed-descriptor defaults (e.g., persistence images) with automated, dataset-adaptive selection from a diverse set of 15 descriptors. The breadth of the underlying 26-dataset evaluation and the agentic architecture with accumulated memory represent concrete strengths that could enable reproducible, generalizable automation if empirically validated.

major comments (2)
  1. [Abstract] Abstract (paragraph beginning 'To fill this gap...'): The central claim that TopoAgent 'analyzes input images and their topological characteristics, reasons about which topological descriptors best suit the input, and determines the optimal descriptor and its configuration, all without task-specific training' is presented without any quantitative results, success rates, ablation studies, or downstream-task metrics on the 26 datasets or held-out data. This absence directly undermines assessment of whether the Perception–Reasoning–Action–Reflection loop with 21 tools actually succeeds at reliable selection.
  2. [Framework description] Description of skill distillation (paragraph on the 26-dataset evaluation): The manuscript states that the agent's skill set 'is distilled from systematic evaluation of 15 topological descriptors across 26 datasets,' yet supplies no tables, figures, or reported metrics (e.g., selection accuracy, comparison to fixed baselines, or generalization tests) showing how the evaluation outcomes translate into verifiable agent behavior on new images. This is load-bearing for the claim of autonomous, training-free operation.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We sincerely thank the referee for the constructive review and recommendation for major revision. The comments highlight important aspects of how results are presented, and we address each point below with plans to strengthen the manuscript.

read point-by-point responses

  1. Referee: [Abstract] Abstract (paragraph beginning 'To fill this gap...'): The central claim that TopoAgent 'analyzes input images and their topological characteristics, reasons about which topological descriptors best suit the input, and determines the optimal descriptor and its configuration, all without task-specific training' is presented without any quantitative results, success rates, ablation studies, or downstream-task metrics on the 26 datasets or held-out data. This absence directly undermines assessment of whether the Perception–Reasoning–Action–Reflection loop with 21 tools actually succeeds at reliable selection.

    Authors: The abstract is a concise summary and does not include specific metrics, which is standard practice. The full manuscript reports quantitative results supporting the claim in the Experiments section, including agent success rates on descriptor selection, ablation studies on the loop components, and downstream performance metrics across the 26 datasets and held-out cases. In revision we will add 1-2 key quantitative highlights (e.g., selection accuracy and average downstream improvement) to the abstract for immediate visibility. revision: yes

  2. Referee: [Framework description] Description of skill distillation (paragraph on the 26-dataset evaluation): The manuscript states that the agent's skill set 'is distilled from systematic evaluation of 15 topological descriptors across 26 datasets,' yet supplies no tables, figures, or reported metrics (e.g., selection accuracy, comparison to fixed baselines, or generalization tests) showing how the evaluation outcomes translate into verifiable agent behavior on new images. This is load-bearing for the claim of autonomous, training-free operation.

    Authors: The underlying evaluation of the 15 descriptors is performed and referenced in the manuscript, with the distilled skills encoded in the agent's tools and memory. We acknowledge that the main text would benefit from an explicit summary of the evaluation metrics. In the revised manuscript we will insert a concise table (or figure) summarizing key outcomes such as per-descriptor performance, selection accuracy on held-out data, and comparison to fixed baselines, directly linking the evaluation to the agent's autonomous behavior. revision: yes

Circularity Check

0 steps flagged

No circularity: framework proposal with no derivations or fitted predictions

full rationale

The paper proposes an LLM-based agentic software framework (TopoAgent) for descriptor selection in TDA. No equations, first-principles derivations, or statistical predictions appear in the provided text. The skill set is informed by prior evaluation on 26 datasets, but this is standard empirical grounding for a tool, not a reduction of any claimed result to its own inputs by construction. No self-citations, uniqueness theorems, or ansatzes are invoked as load-bearing. The contribution is architectural rather than predictive, so none of the enumerated circularity patterns apply.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 1 invented entities

The central claim rests on the domain assumption that topological descriptors derived from persistent homology are useful for medical images and that an LLM can be guided by tools to select among them effectively. No free parameters or invented physical entities are described.

axioms (2)
  • domain assumption Topological descriptors from persistent homology capture geometric structural properties in medical images that conventional pixel-level deep learning approaches often neglect.
    Stated directly in the opening sentence of the abstract as the motivation for the work.
  • ad hoc to paper An LLM agent equipped with domain-specific tools can reason about topological characteristics of input images without task-specific training.
    This is the core operational premise required for the agent to function as described.
invented entities (1)
  • TopoAgent framework no independent evidence
    purpose: Automate selection and configuration of topological descriptors via an LLM agent
    The framework itself is the novel constructed system; no independent evidence outside the proposal is provided.

pith-pipeline@v0.9.1-grok · 5772 in / 1449 out tokens · 31540 ms · 2026-06-30T06:30:11.742779+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

75 extracted references · 11 canonical work pages · 7 internal anchors

  1. [1]

    In: CBMS

    Adame, D., Nunez, J.A., Vazquez, F., Gurrola, N., Li, H., Tang, H., Fu, B., Gu, P.: Topo-VM-UNetV2: Encoding topology into vision Mamba UNet for polyp seg- mentation. In: CBMS. pp. 258–263 (2025)

    2025

  2. [2]

    Journal of Machine Learning Research18(8), 1–35 (2017)

    Adams, H., Emerson, T., Kirby, M., Neville, R., Peterson, C., et al.: Persistence Images: A stable vector representation of persistent homology. Journal of Machine Learning Research18(8), 1–35 (2017)

    2017

  3. [3]

    gradient histograms: A comparative study for medical image classification

    Ahmed, F.: Topological signatures vs. gradient histograms: A comparative study for medical image classification. arXiv preprint arXiv:2507.03006 (2025)

    work page arXiv 2025

  4. [4]

    Journal of Imaging Informatics in Medicine (2025)

    Ahmed, F., Bhuiyan, M.A.N., Coskunuzer, B.: Topo-CNN: Retinal image analysis with topological deep learning. Journal of Imaging Informatics in Medicine (2025)

    2025

  5. [5]

    IEEE Transactions on Pattern Analysis and Machine Intelligence45(12) (2023)

    Ali, D., Asaad, A., Jimenez, M.J., Nanda, V., Paluzo-Hidalgo, E., Soriano- Trigueros, M.: A survey of vectorization methods in topological data analysis. IEEE Transactions on Pattern Analysis and Machine Intelligence45(12) (2023)

    2023

  6. [6]

    anthropic

    Anthropic: The Claude model card (2025),https : / / www . anthropic . com / research/claude-model-card

    2025

  7. [7]

    Asia Pacific Tele-Ophthalmology Society: APTOS 2019 blindness detection (2019), https://www.kaggle.com/c/aptos2019-blindness-detection

    2019

  8. [8]

    Machine Learning45(1), 5–32 (2001)

    Breiman, L.: Random forests. Machine Learning45(1), 5–32 (2001)

    2001

  9. [9]

    The Journal of Machine Learning Research16(1), 77–102 (2015)

    Bubenik, P.: Statistical topological data analysis using persistence landscapes. The Journal of Machine Learning Research16(1), 77–102 (2015)

    2015

  10. [10]

    Bulletin of the American Mathematical Society 46(2), 255–308 (2009)

    Carlsson, G.: Topology and data. Bulletin of the American Mathematical Society 46(2), 255–308 (2009)

    2009

  11. [11]

    In: AISTATS

    Carrière, M., Chazal, F., Ike, Y., Lacombe, T., Royer, M., Umeda, Y.: PersLay: A neuralnetworklayerforpersistencediagramsandnewgraphtopologicalsignatures. In: AISTATS. pp. 2786–2796 (2020)

    2020

  12. [12]

    arXiv preprint arXiv:2507.01171 (2025)

    Chambers, E.W., Meng, G.: A stable and theoretically grounded Gromov- Wasserstein distance for Reeb graph comparison using persistence images. arXiv preprint arXiv:2507.01171 (2025)

    work page arXiv 2025

  13. [13]

    In: ACM SIGKDD

    Chen, T., Guestrin, C.: XGBoost: A scalable tree boosting system. In: ACM SIGKDD. pp. 785–794 (2016)

    2016

  14. [14]

    PLoS ONE10(10), e0140381 (2015)

    Cheng, J., Huang, W., Cao, S., Yang, R., Yang, W., Yun, Z., Wang, Z., Feng, Q.: Enhanced performance of brain tumor classification via tumor region augmentation and partition. PLoS ONE10(10), e0140381 (2015)

    2015

  15. [15]

    The Lancet Digital Health4(5), e330–e339 (2022)

    Combalia, M., Codella, N.C., Rotemberg, V., Helba, B., Vilaplana, V., Reiter, O., Carrera, C., Barreiro, A., Halpern, A.C., Puig, S., Malvehy, J.: Validation of artificial intelligence prediction models for skin cancer diagnosis using dermoscopy images: The 2019 International Skin Imaging Collaboration Grand Challenge. The Lancet Digital Health4(5), e330–...

    2019

  16. [16]

    Discrete & Computational Geometry28(4), 511–533 (2002)

    Edelsbrunner, Letscher, Zomorodian: Topological persistence and simplification. Discrete & Computational Geometry28(4), 511–533 (2002)

    2002

  17. [17]

    TabArena: A Living Benchmark for Machine Learning on Tabular Data

    Erickson, N., Purucker, L., Tschalzev, A., Holzmüller, D., Desai, P.M., Salinas, D., Hutter, F.: TabArena: A living benchmark for machine learning on tabular data. arXiv preprint arXiv:2506.16791 (2025)

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  18. [18]

    arXiv preprint arXiv:2502.02673 (2025)

    Fallahpour, A., Ma, J., Munim, A., Lyu, H., Wang, B.: MedRAX: Medical reason- ing agent for chest X-Ray. arXiv preprint arXiv:2502.02673 (2025)

    work page arXiv 2025

  19. [19]

    Advances in Neural Information Processing Systems28(2015) TopoAgent for Automated Topology Learning 17

    Feurer, M., Klein, A., Eggensperger, K., Springenberg, J., Blum, M., Hutter, F.: Efficient and robust automated machine learning. Advances in Neural Information Processing Systems28(2015) TopoAgent for Automated Topology Learning 17

    2015

  20. [20]

    Google DeepMind: Gemini 2.5: Our most intelligent AI model (2025),https:// deepmind.google/technologies/gemini/

    2025

  21. [21]

    In: ICLR (2025)

    Gorishniy, Y., Kotelnikov, A., Babenko, A.: TabM: Advancing tabular deep learn- ing with parameter-efficient ensembling. In: ICLR (2025)

    2025

  22. [22]

    The Llama 3 Herd of Models

    Grattafiori, A., Dubey, A., Jauhri, A., Pandey, A., Kadian, A., Al-Dahle, A., Let- man,A.,Mathur,A.,Schelten,A.,Vaughan,A.,etal.:TheLlama3herdofmodels. arXiv preprint arXiv:2407.21783 (2024)

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  23. [23]

    In: WACV

    Gu, P., Li, H., Tang, H., Xu, D., Enriquez, E., Kim, D., Fu, B., Chen, D.Z.: Integrating multi-scale and multi-filtration topological features for medical image classification. In: WACV. pp. 8660–8669 (2026)

    2026

  24. [24]

    In: ACM MM

    Gu, P., Wang, H., Zhang, Y., Li, H., Wang, C., Chen, D.: TopoImages: Incorpo- rating local topology encoding into deep learning models for medical image classi- fication. In: ACM MM. pp. 1938–1947 (2025)

    1938

  25. [25]

    In: CVPR

    He, K., Zhang, X., Ren, S., Sun, J.: Deep residual learning for image recognition. In: CVPR. pp. 770–778 (2016)

    2016

  26. [26]

    Nature (2025)

    Hollmann, N., Müller, S., Purucker, L., Krishnakumar, A., Körfer, M., et al.: Ac- curate predictions on small data with a tabular foundation model. Nature (2025)

    2025

  27. [27]

    In: NeurIPS (2024)

    Holzmüller, D., Grinsztajn, L., Obst, N.: Better by default: Strong pre-tuned MLPs and boosted trees on tabular data. In: NeurIPS (2024)

    2024

  28. [28]

    GPT-4o System Card

    Hurst, A., Lerer, A., Goucher, A.P., Perelman, A., Ramesh, A., Clark, A., et al.: GPT-4o system card. arXiv preprint arXiv:2410.21276 (2024)

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  29. [29]

    Springer (2019)

    Hutter, F., Kotthoff, L., Vanschoren, J.: Automated Machine Learning: Methods, Systems, Challenges. Springer (2019)

    2019

  30. [30]

    IEEE Access (2025)

    Khamis, A.: Agentic AI systems: Architecture and evaluation using a frictionless parking scenario. IEEE Access (2025)

    2025

  31. [31]

    Advances in Neural Information Processing Systems33, 15965–15977 (2020)

    Kim, K., Kim, J., Zaheer, M., Kim, J., Chazal, F., Wasserman, L.: PLLay: Efficient topological layer based on persistent landscapes. Advances in Neural Information Processing Systems33, 15965–15977 (2020)

    2020

  32. [32]

    LangChain, Inc.: LangGraph: Build resilient language agents as graphs (2024), https://github.com/langchain-ai/langgraph

    2024

  33. [33]

    Scientific Reports9(1), 1139 (2019)

    Lawson,P.,Sholl, A.B.,Brown,J.Q., Fasy, B.T.,Wenk,C.: Persistenthomology for the quantitative evaluation of architectural features in prostate cancer histology. Scientific Reports9(1), 1139 (2019)

    2019

  34. [34]

    Scientific Data4(1), 1–9 (2017)

    Lee, R.S., Gimenez, F., Hoogi, A., Miyake, K.K., Gorovoy, M., Rubin, D.L.: A curated mammography data set for use in computer-aided detection and diagnosis research. Scientific Data4(1), 1–9 (2017)

    2017

  35. [35]

    In: Findings of EMNLP

    Li, B., Yan, T., Pan, Y., Luo, J., Ji, R., et al.: MMedAgent: Learning to use medical tools with multi-modal agent. In: Findings of EMNLP. pp. 8745–8760 (2024)

    2024

  36. [36]

    IEEE Transactions on Medical Imaging45(4), 1311–1324 (2026)

    Liang, P., Ding, Y., Zhang, Y., Chen, J., Zheng, H., Wang, H., Zhang, Y., Meng, G., Weninger, T., Niemier, M., et al.: Cell instance segmentation: The devil is in the boundaries. IEEE Transactions on Medical Imaging45(4), 1311–1324 (2026)

    2026

  37. [37]

    In: CVPR

    Lin, J., Zhu, C., Kneuertz, P.J., Bai, Y., Xue, Y.: When models learn to ask why: Adaptive causal reasoning for trustworthy medical vision-language models. In: CVPR. pp. 5556–5568 (2026)

    2026

  38. [38]

    Foundations of Data Science6(2), 221 (2024)

    Liu, J., Shen, L., Wei, G.W.: ChatGPT for computational topology. Foundations of Data Science6(2), 221 (2024)

    2024

  39. [39]

    In: CVPR

    Liu, Z., Hu, H., Lin, Y., Yao, Z., Xie, Z., Wei, Y., Ning, J., Cao, Y., Zhang, Z., Dong, L., et al.: Swin Transformer V2: Scaling up capacity and resolution. In: CVPR. pp. 12009–12019 (2022)

    2022

  40. [40]

    In: ICMS

    Maria, C., Boissonnat, J.D., Glisse, M., Yvinec, M.: GUDHI: Simplicial complexes and persistent homology packages. In: ICMS. pp. 167–174 (2014) 18 G. Meng et al

    2014

  41. [41]

    In: SP-SS, pp

    Mecke, K.R.: Additivity, convexity, and beyond: Applications of Minkowski func- tionals in statistical physics. In: SP-SS, pp. 111–184. Springer (2000)

    2000

  42. [42]

    In: CVPR

    Meng, G., Gu, P., Liang, P., Lalor, J.P., Chambers, E.W., Chen, D.Z.: TopoCL: Topological contrastive learning for medical imaging. In: CVPR. pp. 42681–42690 (2026)

    2026

  43. [43]

    Computational Linguistics pp

    Meng, G., Zeng, Q., Lalor, J.P., Yu, H.: A psychology-based unified dynamic frame- work for curriculum learning. Computational Linguistics pp. 1–49 (2025)

    2025

  44. [44]

    IEEE Transactions on Multimedia (2025)

    Meng, G., Zhou, R., Liu, L., Liang, P., Liu, F., Chen, D.Z., Niemier, M., Hu, X.S.: Efficient approximation of Earth Mover’s distance based on nearest neighbor search. IEEE Transactions on Multimedia (2025)

    2025

  45. [45]

    IEEE Transactions on Pattern Analysis and Machine Intelligence24(7), 971–987 (2002)

    Ojala, T., Pietikainen, M., Maenpaa, T.: Multiresolution gray-scale and rotation invariant texture classification with local binary patterns. IEEE Transactions on Pattern Analysis and Machine Intelligence24(7), 971–987 (2002)

    2002

  46. [46]

    In: WACV

    Peng, Y., Wang, H., Sonka, M., Chen, D.Z.: PHG-Net: Persistent homology guided medical image classification. In: WACV. pp. 7583–7592 (2024)

    2024

  47. [47]

    Foundations of Computational Mathematics23(4), 1215–1272 (2023)

    Perea, J.A., Munch, E., Khasawneh, F.A.: Approximating continuous functions on persistence diagrams using template functions. Foundations of Computational Mathematics23(4), 1215–1272 (2023)

    2023

  48. [48]

    Advances in Neural Information Pro- cessing Systems31(2018)

    Prokhorenkova, L., Gusev, G., Vorobev, A., Dorogush, A.V., Gulin, A.: CatBoost: Unbiased boosting with categorical features. Advances in Neural Information Pro- cessing Systems31(2018)

    2018

  49. [49]

    In: AISTATS (2021)

    Royer, M., Chazal, F., Levrard, C., Umeda, Y., Ike, Y.: ATOL: Measure vector- ization for automatic topologically-oriented learning. In: AISTATS (2021)

    2021

  50. [50]

    Advances in Neural Information Processing Systems36, 8634–8652 (2023)

    Shinn, N., Cassano, F., Gopinath, A., Narasimhan, K., Yao, S.: Reflexion: Lan- guage agents with verbal reinforcement learning. Advances in Neural Information Processing Systems36, 8634–8652 (2023)

    2023

  51. [51]

    Cancer Research (2026)

    Singhal, S., Li, C., Aukerman, A., Carrière, M., Miller, M.L., Hibshoosh, H., Mc- Donald, J.A., Winfield, J.R., et al.: Topology-based biomarkers accurately predict breast cancer outcome and survival. Cancer Research (2026)

    2026

  52. [52]

    IEEE Transactions on Biomedical Engineer- ing63(7), 1455–1462 (2016)

    Spanhol, F.A., Oliveira, L.S., Petitjean, C., Heutte, L.: A dataset for breast cancer histopathological image classification. IEEE Transactions on Biomedical Engineer- ing63(7), 1455–1462 (2016)

    2016

  53. [53]

    Artificial Intelligence Review (2025)

    Su, Z., Liu, X., Hamdan, L.B., Maroulas, V., Wu, J., Carlsson, G., Wei, G.W.: Topological data analysis and topological deep learning beyond persistent homol- ogy: A review. Artificial Intelligence Review (2025)

    2025

  54. [54]

    EndoCogniAgent: Closed-Loop Agentic Reasoning with Self-Consistency Validation for Endoscopic Diagnosis

    Tang, Y., Wang, K., Chen, Y., Zhou, G.: EndoAgent: A memory-guided reflec- tive agent for intelligent endoscopic vision-to-decision reasoning. arXiv preprint arXiv:2508.07292 (2025), withdrawal

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  55. [55]

    Information and Inference: A Journal of the IMA3(4), 310–344 (2014)

    Turner, K., Mukherjee, S., Boyer, D.M.: Persistent homology transform for mod- eling shapes and surfaces. Information and Inference: A Journal of the IMA3(4), 310–344 (2014)

    2014

  56. [56]

    Information and Media Technologies12, 228–239 (2017)

    Umeda, Y.: Time series classification via topological data analysis. Information and Media Technologies12, 228–239 (2017)

    2017

  57. [57]

    Voyager: An Open-Ended Embodied Agent with Large Language Models

    Wang, G., Xie, Y., Jiang, Y., Mandlekar, A., thers: Voyager: An open-ended em- bodied agent with large language models. arXiv preprint arXiv:2305.16291 (2023)

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  58. [58]

    IEEE Transactions on Medical Imaging (2023)

    Wang, H., Huang, G., Zhao, Z., Cheng, L., et al.: CCF-GNN: A unified model ag- gregating appearance, microenvironment, and topology for pathology image clas- sification. IEEE Transactions on Medical Imaging (2023)

    2023

  59. [59]

    Frontiers of Computer Science18(6), 186345 (2024) TopoAgent for Automated Topology Learning 19

    Wang, L., Ma, C., Feng, X., Zhang, Z., Yang, H., Zhang, J., Chen, Z., Tang, J., Chen, X., Lin, Y., et al.: A survey on large language model based autonomous agents. Frontiers of Computer Science18(6), 186345 (2024) TopoAgent for Automated Topology Learning 19

    2024

  60. [60]

    Medagent-pro: Towards evidence-based multi-modal medical diagnosis via reasoning agentic workflow.arXiv preprint arXiv:2503.18968, 2025

    Wang, Z., Wu, J., Cai, L., Low, C.H., Yang, X., Li, Q., Jin, Y.: MedAgent-Pro: Towards evidence-based multi-modal medical diagnosis via reasoning agentic work- flow. arXiv preprint arXiv:2503.18968 (2025)

    work page arXiv 2025

  61. [61]

    Science China Information Sciences68(2), 121101 (2025)

    Xi, Z., Chen, W., Guo, X., He, W., Ding, Y., Hong, B., Zhang, M., Wang, J., Jin, S., Zhou, E., et al.: The rise and potential of large language model based agents: A survey. Science China Information Sciences68(2), 121101 (2025)

    2025

  62. [62]

    In: Proceedings of the AAAI Conference on Artificial Intelligence

    Xiao, J., Wu, Z., Lin, H., Chen, Y., Liu, Y., Zhao, X., Wang, Z., He, Z.: Not just what’sthere:Enablingcliptocomprehendnegatedvisualdescriptionswithoutfine- tuning. In: Proceedings of the AAAI Conference on Artificial Intelligence. vol. 40, pp. 10978–10986 (2026)

    2026

  63. [63]

    Authorea Preprints (2025)

    Xu, G., Li, X., Chen, Y., Duan, Y., Wu, S., Yu, A., Chiu, C.H., et al.: A compre- hensive survey of agentic AI in healthcare. Authorea Preprints (2025)

    2025

  64. [64]

    Scientific Data10(1), 41 (2023)

    Yang, J., Shi, R., Wei, D., Liu, Z., Zhao, L., Ke, B., Pfister, H., Ni, B.: MedM- NIST v2–a large-scale lightweight benchmark for 2D and 3D biomedical image classification. Scientific Data10(1), 41 (2023)

    2023

  65. [65]

    In: ICLR (2022)

    Yao, S., Zhao, J., Yu, D., Du, N., Shafran, I., Narasimhan, K.R., Cao, Y.: ReAct: Synergizing reasoning and acting in language models. In: ICLR (2022)

    2022

  66. [66]

    Yu, J., Xiao, J., Zhu, C., Li, J., Li, J., Deng, H., Wang, X., Jia, G., Li, J., Bai, X., Zhou, B., Ma, Z.: I2e: From image pixels to actionable interactive environments for text-guided image editing (2026),https://arxiv.org/abs/2601.03741

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  67. [67]

    In: BIBM

    Zhu, C., Lin, J., Tan, G., Zhu, N., Li, K., Wang, C., Li, S.: Advancing ultrasound medical continuous learning with task-specific generalization and adaptability. In: BIBM. pp. 3019–3025 (2024)

    2024

  68. [68]

    In: AAAI

    Zhu, C., Lin, Y., Chen, S., Wang, Y., Lin, J.: MedEyes: Learning dynamic visual focus for medical progressive diagnosis. In: AAAI. vol. 40, pp. 13916–13924 (2026)

    2026

  69. [69]

    In: ACM MM

    Zhu, C., Lin, Y., Shao, J., Lin, J., Wang, Y.: Pathology-aware prototype evolu- tion via LLM-driven semantic disambiguation for multicenter diabetic retinopathy diagnosis. In: ACM MM. pp. 9196–9205 (2025)

    2025

  70. [70]

    IEEE Transactions on Circuits and Systems for Video Technology (2025)

    Zhu, C., Shao, J., Lin, J., Wang, Y., Wang, J., Tang, J., Li, K.: fMRI2GES: Co- speech gesture reconstruction from fMRI signal with dual brain decoding align- ment. IEEE Transactions on Circuits and Systems for Video Technology (2025)

    2025

  71. [71]

    In: MICCAI (2026)

    Zhu, C., Wang, Y., Lin, J., Wang, F., Wang, H., Zhao, L., Li, S., Li, K.: Anatomy- anchored self-supervision: Distilling vision foundation models for invariant ultra- sound representation. In: MICCAI (2026)

    2026

  72. [72]

    MedSynapse-V: Bridging Visual Perception and Clinical Intuition via Latent Memory Evolution

    Zhu, C., Zeng, J., Jiang, J., Lin, J., Wang, Y.: Medsynapse-v: Bridging visual perception and clinical intuition via latent memory evolution. arXiv preprint arXiv:2604.26283 (2026)

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  73. [73]

    In: CVPR

    Zhu, W., Zhang, Y., Jin, X., Zeng, W., Zhang, L.: ANTS: Adaptive negative tex- tual space shaping for OOD detection via test-time MLLM understanding and reasoning. In: CVPR. pp. 20–30 (2026)

    2026

  74. [74]

    ISIC Challenge (2018)

    Zhuang, J., Li, W., Manivannan, S., Wang, R., et al.: Skin lesion analysis towards melanoma detection using deep neural network ensemble. ISIC Challenge (2018)

    2018

  75. [75]

    Artificial Intelligence Review57(4), 77 (2024)

    Zia, A., Khamis, A., Nichols, J., Tayab, U.B., Hayder, Z., Rolland, V., Stone, E., Petersson, L.: Topological deep learning: A review of an emerging paradigm. Artificial Intelligence Review57(4), 77 (2024)

    2024