Pith. sign in

REVIEW 2 major objections 5 minor 35 references

GeoXplain: On-the-Fly Visual Explanations for Weather Foundation Models

T0 review · 2 major / 5 minor · reviewed 2026-07-11 · grok-4.5

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

Pith's one-line read GeoXplain lets analysts interactively inspect weather foundation-model attributions as maps, pressure stacks, and timelines without leaving the notebook.

desk verdict Solid open systems paper: a real notebook/HPC-friendly geospatial XAI viewer plus a working Aurora adapter; usefulness of the attributions is left open, but that is not load-bearing for the toolkit claim. read the letter →

arxiv 2607.05655 v1 pith:MWHSXXUV submitted 2026-07-06 cs.HC

Clemens Walter Koprolin , Leonardo Trentini , Benedikt Soja , Mennatallah El-Assady , Christina Humer This is my paper · ORCID
classification cs.HC
keywords climatevisualizationgeospatialexplainableAIweatherfoundationmodelscomputationalnotebooksattributionmapsAurora
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

Weather foundation models forecast many variables across the globe, pressure levels, and lead times, but the learned influences behind a single forecast are hard to inspect with static plots. GeoXplain is a Python toolkit that treats those influences as geospatial attribution maps: dense grids packaged with metadata so a notebook widget or browser can render map and globe views, switch variables and levels, scrub forecast time, annotate targets, and overlay physical fields such as humidity or wind. Computation is split from display through a self-describing result-bundle format and model adapters. The first adapter runs saliency, Integrated Gradients, RISE, and ViT-CX (plus multi-step rollouts and ERA5 overlays) for the Aurora model, either on a local GPU or through remote GPU or SLURM listeners, while keeping the same analyst call site. The paper argues this on-the-fly, geophysically contextual workflow is what atmospheric scientists need to compare explanations with meteorological structure.

What carries the argument

The result-bundle protocol: attribution grids plus metadata (method, target, timestamps, variables, pressure layers) that fully decouple the viewer from any model-specific adapter and power linked map, globe, timeline, and overlay views.

What would settle it

A controlled study with weather or climate scientists showing they cannot form stable, method-consistent interpretations of model behavior when using GeoXplain on real cases, or diagnostics showing the adapted methods fail completeness, randomization, or physical-consistency checks on the targets they claim to explain.

Watch

Extended reading notes

Core claim

GeoXplain is a model-agnostic earth-system visualization toolkit, paired with an Aurora computation adapter, that makes dense geospatial attribution maps inspectable on demand across variables, pressure levels, and forecast time through a shared result-bundle protocol and notebook or browser interfaces.

Load-bearing premise

That the Aurora-specific versions of the attribution methods produce maps that are scientifically useful for meteorological interpretation, not only implementable and able to pass sanity checks.

Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 5 minor

Summary. GeoXplain is presented as a model-agnostic interactive visualization toolkit for dense geospatial attribution maps from weather and climate foundation models, spanning variables, pressure levels, and forecast time. The system separates a reusable viewer (notebook widget and browser; map/globe modes, timelines, overlays) from model-specific adapters via a self-describing result-bundle protocol. The GeoXplain Aurora Adapter is the first backend: it adapts Saliency, Integrated Gradients, RISE, and ViT-CX (plus multi-frame/rollout modes and ERA5 overlays), runs locally or via GPU/SLURM listeners under a single Python call site, and is released as open-source PyPI packages. Supporting material includes design goals and a usage scenario (§3), adapter architecture (§4), visual-parity and protocol tests (Appendix C), and numerical sanity checks such as IG completeness and parameter randomization (Appendix B).

Significance. If the systems claims hold—as the open packages, narrow adapter interface, remote execution path, and verification suites suggest—this is a practical contribution for earth-system XAI workflows that currently split notebooks, HPC jobs, and static plots. Strengths that should be credited explicitly include: (i) a clean viewer/adapter split with a documented result protocol; (ii) on-the-fly notebook integration with remote GPU/SLURM backends; (iii) open PyPI release and reproducible diagnostic scripts; and (iv) method adaptations and sanity checks (IG completeness, Adebayo-style randomization) rather than visual plausibility alone. The work is more systems/tooling than a new meteorological finding; its value is enabling inspection and comparison of attributions in geophysical context for Aurora and, by design, other models that implement the same bundle interface.

major comments (2)
  1. [§1, §3.2, §6] §1 contribution (1) and §3.2 design goals G1–G4 claim a visualization workflow that preserves uninterrupted notebook analysis and weather-specific context. The manuscript supports this with architecture, a walkthrough scenario (§3.3), and technical verification (§5; Appendices B–C), but provides no user study, expert interview, or even lightweight task-based feedback with atmospheric scientists. §6 defers evaluation. For a cs.HC visualization contribution, that gap is load-bearing for the design claims (as opposed to the toolkit-existence claim). Please either (a) add preliminary expert evaluation of G1–G4, or (b) reframe contribution (1) and the design-goal language as design rationale/system description, and state clearly that workflow benefits are hypothesized pending evaluation.
  2. [Abstract; §1; Appendix A; §3.3; §5] Appendix A documents substantial Aurora-specific choices (Gaussian-smoothed IG baseline; t1-only export; column-wise atmospheric masks for RISE/ViT-CX; encoder-stage clustering; rollout attributions w.r.t. the original input). §3.3 and §5 correctly warn that visual agreement is not physical mechanism and that diagnostics are necessary but incomplete. The Abstract and §1 still risk being read as implying scientifically ready meteorological explanations. Please tighten the framing so that the primary claim is the visualization protocol and adapter infrastructure, and mark the method adaptations as implementable, sanity-checked baselines whose meteorological utility remains open—consistent with the paper’s own caveats.
minor comments (5)
  1. [Fig. 1; Fig. 3] Fig. 1 and Fig. 3 are central to the usage story; ensure final production figures remain readable at column width (control labels 3–9 in Fig. 1; contour vs overlay contrast in Fig. 3).
  2. [§5; Table 2; Table 3] Table 3 reports IG completeness for a useful but still narrow case set; a one-sentence pointer in §5 to the full case list in Table 2 would help readers who skip the appendix.
  3. [§4.3] §4.3 sketches adapting other models well; a short checklist or minimal interface sketch (target resolver, batch builder, method runner, serializer) in the main text or a box would make the model-agnostic claim easier to act on.
  4. [Appendix A] Notation in Appendix A (S(X), τ, X′, δ, etc.) is clear; define n (IG steps) and N (RISE masks) once in a single parameter table reference early in the appendix for skimming readers.
  5. [Abstract; throughout] Minor polish: ensure consistent hyphenation of “on-the-fly,” and check spacing around citations and “ViT-CX” vs “ViT -CX” in the Abstract.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: systems toolkit paper with no derivation that reduces to fitted inputs or self-citation chains.

full rationale

GeoXplain is an engineering/systems paper whose central claims are architectural: a model-agnostic viewer, a self-describing result-bundle protocol, and a working Aurora adapter that computes and packages attribution grids for interactive geospatial display (Abstract; §1 contributions (1)–(5); §3–4). There is no first-principles derivation, uniqueness theorem, or quantitative prediction whose output is forced by construction from fitted parameters. Attribution methods (Saliency, IG, RISE, ViT-CX, rollouts) are standard techniques adapted for Aurora’s input structure (Appendix A); the adaptations are implementation choices (smoothed baselines, column-wise masks, encoder-stage clustering, t1 export), not circular definitions of the claimed result. Verification uses external sanity practices—IG completeness against S(X)−S(X′), Adebayo-style parameter randomization, RISE Monte Carlo convergence, and visual-parity checks against independent Matplotlib references (Appendices B–C)—rather than self-referential fits. Self-description of the software and open-source availability do not constitute circular scientific prediction. The authors explicitly treat diagnostics as necessary but incomplete and defer user evaluation (§5–6), so residual scientific usefulness questions are outside the circularity criterion. Score 0 is therefore the correct outcome.

Assumptions & free parameters 4 free parameters · 4 assumptions · 2 invented entities

Load-bearing content is mostly engineering design plus standard XAI and weather-model assumptions, not free physical constants. Free parameters are method/runtime knobs that affect maps. Axioms are domain practices (scalar targets, grid structure, usefulness of attribution). Invented entities are software abstractions (bundle format, adapter/listener architecture) introduced to make the system work; they have independent evidence only insofar as the open code runs.

free parameters (4)
  • IG integration steps n and Gaussian baseline σ = σ=2.5°; n up to 64 in diagnostics
    Path integral uses midpoint samples (diagnostics use 4–64 steps; baseline σ=2.5°) chosen by implementers; completeness and map appearance depend on these knobs (Appendix A, Table 1).
  • RISE mask resolution, mask count, and seeds = 18×36 grid; N varies
    18×36 low-res Bernoulli masks, upsampling, circular longitude shifts, and N masks control variance and cost; convergence checked at 32–1024 masks (Appendix A–B).
  • ViT-CX hook_stage, n_clusters, optional smoothing = stage 2; K=n_clusters
    Aurora has no single patch grid; stage-2 4×45×90 tokens and clustering hyperparameters define occlusion regions and thus maps (Appendix A, Table 1).
  • Attribution export convention (t1 only; column-wise atmospheric masks)
    Authors export attributions w.r.t. second input step t1 after preliminary testing for interpretability; RISE/ViT-CX collapse atmospheric levels column-wise—design choices that shape all visualized fields (Appendix A).
assumptions (4)
  • domain assumption A user-specified forecast target can be reduced to a differentiable scalar (point value or box mean) suitable for attribution.
    All runners explain S(X)=τ(F(X)) with point or box-mean τ (§3.3, Appendix A).
  • domain assumption Standard attribution methods (saliency, IG, RISE, ViT-CX), after Aurora-specific adaptation, are appropriate objects for interactive meteorological inspection.
    Central product is visualization of these maps; scientific validity is not proven and is explicitly limited (§2, §5, Appendix B).
  • ad hoc to paper Dense regular lat–lon grids (Aurora/WeatherBench-style) are the primary data model for the protocol.
    Protocol optimized for regular grids; unstructured meshes, stations, ensembles deferred (§6).
  • ad hoc to paper Separating a model-agnostic viewer from adapters preserves uninterrupted notebook workflows while supporting remote GPU/SLURM execution.
    Design goals G1–G4 and architecture in §3–4; user benefit not yet evaluated (§6).
invented entities (2)
  • GeoXplain result / overlay bundle protocol (.xai.npz, .overlay.npz, metadata schema) independent evidence
    purpose: Decouple visualization from model-specific computation so any adapter can feed the same viewer.
    Core contribution (2) and Appendix E; independent evidence is the open implementation, not external theory.
  • GeoXplain Aurora Adapter with local/GPU-listener/SLURM backends independent evidence
    purpose: Compute on-the-fly attributions and ERA5 overlays for Aurora under a single Python call site.
    §4 and Appendix F; evidence is shipped code and diagnostics.

how reviews work

0 comments
Cite this review

Pith. "Pith review of GeoXplain: On-the-Fly Visual Explanations for Weather Foundation Models." pith.science (2026). https://pith.science/paper/MWHSXXUV

@misc{pith2026260705655,
  author       = {Pith},
  title        = {Pith review of: GeoXplain: On-the-Fly Visual Explanations for Weather Foundation Models},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MWHSXXUV}},
  note         = {Machine review of arXiv:2607.05655}
}
read the original abstract

Weather and climate foundation models produce high-dimensional forecasts whose learned relationships are difficult to inspect with static plots alone. GeoXplain is an interactive Python-based visualization toolkit for exploring geospatial attribution maps across climate variables, atmospheric pressure levels, and forecast time. The toolkit accepts attribution bundles containing attribution grids together with corresponding metadata and renders them in a notebook widget or browser with map and globe modes, linked timelines, pressure-level controls, target annotations, and optional physical-field overlays. We frame GeoXplain as a model-agnostic earth-system visualization toolkit and present the GeoXplain Aurora Adapter as its first computation backend. The adapter computes explanations for the Aurora foundation model, either in a local GPU process, through a GPU listener, or through a SLURM-backed listener, while preserving the same Python call site for analysts. It currently supports gradient saliency, Integrated Gradients, RISE, ViT-CX, multi-frame saliency and Integrated Gradients rollouts, and retrieval of ERA5 overlays. GeoXplain can be installed as a PyPI package with pip install geoxplain. The code is open-source and available at https://github.com/clemenskoprolin/geoxplain.

Figures

Figures reproduced from arXiv: 2607.05655 by the authors.

Figure 1
Figure 1. Example of GeoXplain notebook for explaining a humidity forecast at 850 hPa over Zurich, Switzerland with Integrated Gradients. [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. GeoXplain separates interactive visual analysis from model-specific computation. In the notebook (1), the analyst specifies the target and [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. The left image shows IG attributions for the 850 hPa humidity forecast over Zurich with an 850 hPa temperature overlay. The right image [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (3 more)
Figure 4
Figure 4. Figure 4: Example output from the viewer visual parity suite. The left panel is the independent Matplotlib reference for the sequential opacity-blob case, [PITH_FULL_IMAGE:figures/full_fig_p009_4.png]
Figure 5
Figure 5. Figure 5: GeoXplain result bundles. Attribution bundles store a manifest plus one array per frame, input variable, and layer. Overlay bundles store one [PITH_FULL_IMAGE:figures/full_fig_p009_5.png]
Figure 6
Figure 6. Figure 6: Remote execution in the Aurora adapter. A notebook client submits an HTTP job, polls the listener for progress and logs, and fetches packed [PITH_FULL_IMAGE:figures/full_fig_p009_6.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

35 extracted references · 18 canonical work pages

  1. [1]

    Adebayo, J

    J. Adebayo, J. Gilmer, M. Muelly, I. Goodfellow, M. Hardt, and B. Kim. Sanity checks for saliency maps. InProc. NeurIPS, pp. 9525–9536, 2018. 2

    work page 2018

  2. [2]

    Baño-Medina, A

    J. Baño-Medina, A. Sengupta, J. D. Doyle, C. A. Reynolds, D. Watson- Parris, and L. Delle Monache. Are AI weather models learning atmo- spheric physics? a sensitivity analysis of cyclone Xynthia.npj Climate and Atmospheric Science, 8(1):92, 2025. doi: 10.1038/s41612-025-00949 -6 2

    work page doi:10.1038/s41612-025-00949 2025

  3. [3]

    M. Beg, J. Taka, T. Kluyver, A. Konovalov, M. Ragan-Kelley, N. M. Thiéry et al. Using Jupyter for reproducible scientific workflows.Computing in Science & Engineering, 23(2):36–46, 2021. doi: 10.1109/MCSE.2021. 3052101 1

    doi:10.1109/mcse.2021 2021

  4. [4]

    Ben Bouallègue, M

    Z. Ben Bouallègue, M. C. A. Clare, L. Magnusson, E. Gascon, M. Maier- Gerber, M. Janousek et al. The rise of data-driven weather forecasting: A first statistical assessment of machine learning-based weather forecasts in an operational-like context.Bulletin of the American Meteorological Society, 105(6):E864–E883, 2024. doi: 10.1175/BAMS-D-23-0162.1 1

    doi:10.1175/bams-d-23-0162.1 2024

  5. [5]

    K. Bi, L. Xie, H. Zhang, X. Chen, X. Gu, and Q. Tian. Accurate medium-range global weather forecasting with 3D neural networks.Na- ture, 619:533–538, 2023. doi: 10.1038/s41586-023-06185-3 1, 2

    doi:10.1038/s41586-023-06185-3 2023

  6. [6]

    Bodnar, W

    C. Bodnar, W. P. Bruinsma, A. Lucic, M. Stanley, A. Allen, J. Brandstetter et al. A foundation model for the Earth system.Nature, 641:1180–1187,

    work page

  7. [7]

    doi: 10.1038/s41586-025-09005-y 1, 2

    doi:10.1038/s41586-025-09005-y

  8. [8]

    P. L. Bommer, M. Kretschmer, A. Hedström, D. Bareeva, and M. M.- C. Höhne. Finding the right XAI method—a guide for the evaluation and ranking of explainable AI methods in climate science.Artificial Intelligence for the Earth Systems, 3(3):e230074, 2024. doi: 10.1175/ AIES-D-23-0074.1 2

    work page 2024

Show all 35 references
  1. [9]

    Diehl, L

    A. Diehl, L. Pelorosso, C. Delrieux, C. Saulo, J. Ruiz, M. E. Gröller et al. Visual analysis of spatio-temporal data: Applications in weather forecasting.Computer Graphics Forum, 34(3):381–390, 2015. doi: 10. 1111/cgf.12650 1

    2015

  2. [10]

    Espeholt, S

    L. Espeholt, S. Agrawal, C. Sønderby, M. Kumar, J. Heek, C. Bromberg et al. Deep learning for twelve hour precipitation forecasts.Nature Communications, 13(1):5145, 2022. doi: 10.1038/s41467-022-32483-x 2

    2022 doi

  3. [11]

    Hohman, M

    F. Hohman, M. Kahng, R. Pienta, and D. H. Chau. Visual analytics in deep learning: An interrogative survey for the next frontiers.IEEE Transactions on Visualization and Computer Graphics, 25(8):2674–2693, 2019. doi: 10 .1109/TVCG.2018.2843369 2

    2019

  4. [12]

    Hohman, H

    F. Hohman, H. Park, C. Robinson, and D. H. Chau. Summit: Scaling deep learning interpretability by visualizing activation and attribution summa- rizations.IEEE Transactions on Visualization and Computer Graphics, 26(1):1096–1106, 2020. doi: 10.1109/TVCG.2019.2934659 2

    2020 doi

  5. [13]

    Höllt, A

    T. Höllt, A. Magdy, P. Zhan, G. Chen, G. Gopalakrishnan, I. Hoteit et al. Ovis: A framework for visual analysis of ocean forecast ensembles.IEEE Transactions on Visualization and Computer Graphics, 20(8):1114–1126,

  6. [14]

    doi: 10.1109/TVCG.2014.2307892 2

    2014 doi

  7. [15]

    Z. Hua, R. A. Sobash, D. J. Gagne, II, Y . Sha, and A. Anderson-Frey. Improving medium-range severe weather prediction through transformer postprocessing of AI weather forecasts.Artificial Intelligence for the Earth Systems, 5(1):250045, 2026. doi: 10.1175/AIES-D-25-0045.1 2

    2026 doi

  8. [16]

    J. D. Hunter. Matplotlib: A 2d graphics environment.Computing in Science & Engineering, 9(3):90–95, 2007. doi: 10.1109/MCSE.2007.55 1

    2007 doi

  9. [17]

    Kahng, P

    M. Kahng, P. Y . Andrews, A. Kalro, and D. H. Chau. ActiVis: Visual exploration of industry-scale deep neural network models.IEEE Transac- tions on Visualization and Computer Graphics, 24(1):88–97, 2018. doi: 10.1109/TVCG.2017.2744718 2

    2018 doi

  10. [18]

    Kasteleyn and A

    E. Kasteleyn and A. Lucic. Does Aurora encode atmospheric structure? latent regime analysis and attribution. InICLR 2026 Workshop on Foun- dation Models for Science, 2026. Poster. 2

    2026

  11. [19]

    R. Lam, A. Sanchez-Gonzalez, M. Willson, P. Wirnsberger, M. Fortunato, F. Alet et al. Learning skillful medium-range global weather forecasting. Science, 382:eadi2336, 2023. doi: 10.1126/science.adi2336 1, 2

    2023 doi

  12. [20]

    Mamalakis, E

    A. Mamalakis, E. A. Barnes, and I. Ebert-Uphoff. Investigating the fidelity of explainable artificial intelligence methods for applications of convolutional neural networks in geoscience.Artificial Intelligence for the Earth Systems, 1(4):e220012, 2022. doi: 10.1175/AIES-D-22-0012.1 2

    2022 doi

  13. [21]

    T. Manz, N. Abdennur, and N. Gehlenborg. anywidget: reusable widgets for interactive analysis and visualization in computational notebooks. Journal of Open Source Software, 9(102):6939, 2024. doi: 10.21105/joss. 06939 2

    2024 doi

  14. [22]

    Nayeem, D

    A.-A.-R. Nayeem, D. Han, H. Lee, D. Kim, D. Feldman, W. J. Tolone et al. A survey on exploratory spatiotemporal visual analytics approaches for climate science, 2024. doi: 10.48550/arXiv.2407.21199 1

    2024 doi arXiv

  15. [23]

    Nguyen, J

    T. Nguyen, J. Brandstetter, A. Kapoor, J. K. Gupta, and A. Grover. ClimaX: A foundation model for weather and climate. InProceedings of the 40th International Conference on Machine Learning, 2023. 2

    2023

  16. [24]

    Ozdemir, Y

    F. Ozdemir, Y . Cheng, S. Mohebi, F. Lehmann, S. Adamov, Z. Zhang et al. Earth system foundation model (esfm): A unified framework for heterogeneous data integration and forecasting, 2026. 2

    2026

  17. [25]

    Pathak, S

    J. Pathak, S. Subramanian, P. Harrington, S. Raja, A. Chattopadhyay, M. Mardani et al. FourCastNet: A global data-driven high-resolution weather model using adaptive Fourier neural operators, 2022. 2

    2022

  18. [26]

    Petsiuk, A

    V . Petsiuk, A. Das, and K. Saenko. RISE: Randomized input sampling for explanation of black-box models. InProc. BMVC, 2018. 2

    2018

  19. [27]

    Price, A

    I. Price, A. Sanchez-Gonzalez, F. Alet, T. R. Andersson, A. El-Kadi, D. Masters et al. Probabilistic weather forecasting with machine learning. Nature, 637(8044):84–90, 2025. doi: 10.1038/s41586-024-08252-9 2

    2025 doi

  20. [28]

    S. Rasp, S. Hoyer, A. Merose, I. Langmore, P. Battaglia, T. Russell et al. WeatherBench 2: A benchmark for the next generation of data-driven global weather models.Journal of Advances in Modeling Earth Systems, 16(6):e2023MS004019, 2024. doi: 10.1029/2023MS004019 2

    2024 doi

  21. [29]

    Rautenhaus, M

    M. Rautenhaus, M. Kern, A. Schäfler, and R. Westermann. Three- dimensional visualization of ensemble weather forecasts – part 1: The visualization tool Met.3D (version 1.0).Geoscientific Model Development, 8(7):2329–2353, 2015. doi: 10.5194/gmd-8-2329-2015 2

    2015 doi

  22. [30]

    Simonyan, A

    K. Simonyan, A. Vedaldi, and A. Zisserman. Deep inside convolutional networks: Visualising image classification models and saliency maps. In Workshop at International Conference on Learning Representations, 2014. 2

    2014

  23. [31]

    Sundararajan, A

    M. Sundararajan, A. Taly, and Q. Yan. Axiomatic attribution for deep networks. InProc. ICML, pp. 3319–3328, 2017. 2

    2017

  24. [32]

    Y . Wu, J. M. Hellerstein, and A. Satyanarayan. B2: Bridging code and interactive visualization in computational notebooks. InProc. ACM Sym- posium on User Interface Software and Technology, pp. 152–165, 2020. doi: 10.1145/3379337.3415851 2

    2020 doi

  25. [33]

    Xie, X.-H

    W. Xie, X.-H. Li, C. C. Cao, and N. L. Zhang. ViT-CX: Causal explanation of vision transformers. InProceedings of the Thirty-Second International Joint Conference on Artificial Intelligence, pp. 1569–1577. International Joint Conferences on Artificial Intelligence Organization...

    2023

  26. [34]

    Y . Yu, L. Shen, F. Long, H. Qu, and H. Chen. PyGWalker: On-the-fly assistant for exploratory visual data analysis. InProc. IEEE VIS, pp. 6–10,

  27. [35]

    doi: 10.1109/VIS55277.2024.00009 1, 2 A METHODRUNNERADAPTATIONS All method runners explain the same scalar target. Let X denote the Aurora input case and let S(X) =τ(F(X)) be the scalar obtained by applying Aurora F and resolving the user target τ: either a point value or an a...

    2024 doi

Pith tools

Reviewed July 11, 2026 · model on record in the stance chip above.