Pith. sign in

REVIEW 4 major objections 3 minor 34 references

Bi-cross validation for estimating spectral clustering hyper parameters

T0 review · 4 major / 3 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read Bi-cross validation on an inverted Laplacian estimates spectral clustering hyperparameters

desk verdict A genuine but unproven BCV-on-Laplacian heuristic that fails on the authors' own close-cluster simulation; worth refereeing, not accepting as validated. read the letter →

arxiv 1908.03747 v3 pith:BRZO2KO6 submitted 2019-08-10 stat.ML cs.LGphysics.acc-ph

Sioan Zohar , Chun-Hong Yoon This is my paper · ORCID
classification stat.MLcs.LGphysics.acc-ph MSC 62H3062H25
keywords bi-crossvalidationspectralclusteringnumberofclustersRBFkernelLaplacianmatrixsingularvaluedecompositionx-rayscatteringdatahyperparameterestimation
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

Bi-cross validation (BCV) was previously used to count how many singular vectors are needed to reconstruct a matrix without fitting noise. This paper shows that the same tool applies to spectral clustering once the normalized Laplacian is regularized and inverted: inverting the Laplacian swaps the smallest eigenvectors that encode clusters into the largest singular vectors, so the BCV minimum selects the cluster count $k$ and the RBF kernel width $\Gamma$ at the same time. The paper demonstrates the procedure on simulated clusters, including nested clusters at two length scales, and on experimental x-ray scattering data, where it separates low-fluence dropped shots from signal without manual thresholds. The upshot is a single loss function for spectral-clustering hyperparameters, treating cluster number and length scale as one decision rather than two.

What carries the argument

The central object is the regularized, inverted normalized Laplacian $L_r^{-1} = (L_n + \xi R)^{-1}$, with $R = H - H^T L_n H$ for a Haar-distributed random matrix $H$; the regularization makes the singular matrix invertible. The work it does: inverting the Laplacian reverses the spectral ordering, so the few smallest eigenvectors that represent clusters become the leading singular vectors, the regime where the BCV rank-selection loss (quadrant hold-out plus $k$-term SVD reconstruction) is meaningful. The minimum of the BCV loss surface $\mathrm{BCV}(k,\Gamma)$ is then read out as the estimated number of clusters and the RBF width together.

What would settle it

Take the close-cluster simulation from Figure 1(d), seven clusters in 2D with reduced inter-cluster spacing, and measure the BCV minimum over $k$ and $\Gamma$ for regularization values $\xi = 10^{-14}$, $6.3 \times 10^{-13}$, $10^{-12}$, and $2.5 \times 10^{-9}$. If the minimum does not occur at $k = 7$ for all $\xi$, or if it changes with the random draw of $R$, the claim that the BCV minimum estimates the cluster count in that regime is falsified.

Watch

Extended reading notes

Core claim

The central claim, captured in equation 7, is that BCV performed on the inverted regularized normalized Laplacian $L_r^{-1}$ estimates the spectral clustering hyperparameters. Because the normalized Laplacian $L_n = I - D^{-1/2} W D^{-1/2}$ is singular, the paper adds a small random regularization term $\xi R$ with $R = H - H^T L_n H$ for Haar-distributed $H$, giving $L_r = L_n + \xi R$. Inverting $L_r$ turns the cluster information carried by the smallest eigenvalues of $L_n$ into the information carried by the largest singular values of $L_r^{-1}$, which is exactly the regime where BCV's leave-out reconstruction loss is designed to work. The BCV score is the averaged squared error of predicting a held-out quadrant of $L_r^{-1}$ from the $k$-term SVD of the remaining block, and its minimum over $k$ and $\Gamma$ is taken to be the estimate of the number of clusters and the RBF width. The mapping is what makes the two hyperparameters estimable from one objective, and the paper argues this is not accidental: the question of how many clusters exist cannot be separated from the length scale at which they are defined.

Load-bearing premise

The whole procedure rests on the assumption that after adding a tiny random matrix and inverting, the leading singular vectors of the regularized inverse still align with the smallest eigenvectors of the original normalized Laplacian, so the BCV minimum reports the true cluster count.

Editorial extensions

If this is right

  • Spectral clustering hyperparameters can be chosen by a single objective: minimizing BCV loss on the inverted Laplacian gives both $k$ and $\Gamma$, removing the need for separate heuristics.
  • In experimental x-ray scattering data, the method labels low-fluence dropped shots and signal shots without manually setting detector-fluence thresholds, so data triage can be automated.
  • Because the matrix BCV formulation does not include a classification step, it avoids the pre-conditioning rotations that earlier BCV-for-k-means approaches needed when clusters are not linearly separable.
  • The loss surface can have multiple meaningful minima: the same dataset can yield estimates of 11 clusters at a short length scale and 3 clusters at a longer scale, making multiscale structure visible.
  • The approach is a step toward identifying rare or anomalous events in large per-pulse datasets, since those events appear as small clusters once the right length scale is selected.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If the spectral inversion mapping is as general as claimed, the same invert-to-turn-eigen-counting-into-singular-rank-selection trick could be applied to other eigengap-based methods, such as community detection or manifold clustering, wherever a singular matrix needs regularizing.
  • One testable extension is to average the BCV loss over several independent Haar draws of $R$; the paper uses one draw per run, and averaging would make the selection criterion less dependent on the random regularization and might sharpen the minimum for close clusters.
  • A proof that the BCV minimum coincides with the true cluster count would likely need a bound on how the singular vectors of $L_r^{-1}$ deviate from the smallest eigenvectors of $L_n$ as $\xi \to 0$; the condition number of $L_r$ could supply such a bound and also a principled choice of $\xi$ in place of the empirically scanned $10^{-9}$ to $10^{-14}$.
  • For terabyte-scale data, the quadratic cost of forming and inverting the full Laplacian is a bottleneck; a natural extension is to approximate the leading singular subspace of $L_r^{-1}$ on a random row and column sample and compute the BCV loss on the approximation, which would test the method's scalability.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 3 minor

Summary. The paper proposes using bi-cross validation (BCV) applied to a regularized, inverted normalized Laplacian matrix to jointly estimate the number of clusters k and the RBF kernel width Γ in spectral clustering. The authors claim that inverting the Laplacian converts the problem of selecting the k smallest eigenvectors into selecting the k largest singular vectors, which BCV can handle. They present synthetic experiments with well-separated and close clusters, an experimental application to LCLS x-ray scattering data, and argue that this approach avoids separate hyperparameter tuning. The central connection is stated without proof, and the paper's own Figure 1(d) shows that the estimated k depends on the arbitrarily chosen regularization parameter ξ.

Significance. If the central claim were established, the paper would offer a practical method for jointly selecting spectral-clustering hyperparameters, which is an important problem for large-scale x-ray data analysis. The experimental demonstration on LCLS data and the connection to rare-event identification are potentially useful. However, the core theoretical mapping from singular vectors of the regularized inverse Laplacian to spectral-clustering eigenvectors is not proven and is contradicted by the authors' own synthetic example in the close-cluster regime. The experimental validation is also limited to a small, non-random subset of the data. These issues currently outweigh the potential significance.

major comments (4)
  1. [Section II, Eqs. (5)-(7)] The central identification of the top singular vectors of Lr^{-1} with the k smallest eigenvectors of the normalized Laplacian is not established. Because R = H - H^T L_n H is generically nonsymmetric, L_r is nonsymmetric, and its SVD left and right singular vectors are not its eigenvectors. No argument shows that the k-dimensional SVD subspace of Lr^{-1} spans the spectral-clustering eigenspace of L_n. The paper's own Figure 1(d) is a direct counterexample: the BCV minimum occurs at k=7 only for ξ=2.5e-9, while for the other tested values (1e-14, 6.3e-13, 1e-12) the minimum is at different k values, so the estimated cluster count depends on the arbitrary regularization parameter.
  2. [Section II, after Eq. (5)] The regularization parameter ξ is described as "empirically determined to be of the order 1e-9 to 1e-14," but no procedure is given for choosing its value, and the method's output is sensitive to this choice, as shown in Figures 1(c), 1(d), and 2. Since the central claim is joint estimation of k and Γ, a method whose output varies with an unspecified, user-chosen ξ is not well-defined as a hyperparameter estimator.
  3. [Section IV] The experimental validation is performed on a non-randomly selected 750-row subset of less than 1% of the available data, with no reported error bars, no repeated trials, and an ad hoc additional feature constructed by appending the degree-matrix diagonal calculated with Γ=1e-2. The authors themselves state that the result does not represent the expected number of clusters for the full data set. This demonstration cannot support the general claim that the method reliably identifies rare events or dropped shots in LCLS data.
  4. [Section V] The Discussion explicitly states that "a mathematical proof demonstrating that the BCV loss function minimum correctly estimates the hyper parameters would have to be shown." Because this proof is exactly the load-bearing component of the paper's central claim, and because the numerical evidence in the close-cluster regime contradicts the claim, the manuscript does not currently establish its main result.
minor comments (3)
  1. [Introduction] There are several typographical errors, including redundant words such as "methods methods," "timing timing," and later "observed observed." These should be corrected.
  2. [Section III, Figure 1 caption] The caption says "clustered into 5 groups" for panel (a) and "clusters in to 11 groups" for Figure 2(a); "in to" should be "into." Also, the text references "scikit-learn version 0.19.1" but no version pinning is provided for the BCV code itself, which would help reproducibility.
  3. [Section II, Eq. (8)] The notation for the submatrices is inconsistent: Eq. (7) labels the bottom-right quadrant as E, but the sentence describing the partition says "A was designated as the hold out and 2x2 BCV was configured such that the sub matrices A,B,C,D have the same number of rows and columns," listing D instead of E. Please clarify the naming.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: BCV is holdout-based; the unproven equivalence to spectral clustering is a validity gap, not a circular reduction.

full rationale

The derivation chain does not reduce to its inputs. The BCV loss is computed on held-out quadrants of the regularized inverted Laplacian (Eq. 8), so the estimated number of clusters and RBF width are not fitted to the true cluster count by construction. The regularization parameter xi is an empirical nuisance parameter, and the paper explicitly reports failure cases where the BCV minimum depends on xi (Fig. 1d), which is evidence of a validity gap rather than circularity. The central assertion that the leading singular vectors of Lr^{-1} correspond to spectral-clustering eigenvectors is unproved and the paper concedes this in Section V ('A mathematical proof demonstrating that the BCV loss function minimum correctly estimates the hyper parameters would have to be shown'), but an unproved assumption is not an input-output equivalence. The self-citations ([7], [12], [29]) are contextual or support a preprocessing step (multiplying photon-energy by intensity) and do not carry the load-bearing mapping. No fitted parameter is renamed as a prediction, and no uniqueness theorem is imported from the authors' prior work.

Assumptions & free parameters 2 free parameters · 5 assumptions · 0 invented entities

The central claim rests on one unproved mapping from inverted Laplacian singular vectors to spectral clustering eigenvectors, an empirically chosen regularization parameter, and an ad hoc feature-engineering step for the experimental data.

free parameters (2)
  • regularization parameter ξ = 1e-9 to 1e-14 (empirically determined)
    Added to make the Laplacian invertible; the BCV minimum and estimated cluster count depend on its value, as shown in Figure 1(d).
  • RBF width Γ for degree-matrix feature = 1e-2
    Used in Section IV to compute the local density feature appended to the experimental data; this manual choice affects the clustering result.
assumptions (5)
  • domain assumption BCV applied to a matrix estimates the number of dominant singular vectors needed for reconstruction (Owen and Perry 2009).
    The method relies on this property when applied to the inverted Laplacian.
  • ad hoc to paper The leading singular vectors of the regularized inverted Laplacian Lr^{-1} correspond to the smallest eigenvectors of the normalized Laplacian used in spectral clustering.
    This mapping is asserted in Section II but not proved; the random regularization R perturbs the spectrum and may break the correspondence.
  • standard math The RBF kernel with parameter Γ captures cluster structure at a meaningful scale (Eq. 2).
    Standard choice in spectral clustering.
  • standard math The matrix R = H - H^T Ln H makes Ln + ξR invertible.
    Relying on the construction to avoid singularity; no proof given in the paper, but plausible for random H.
  • ad hoc to paper Appending the diagonal of the degree matrix to the feature space mitigates heterogeneous density in experimental data.
    Introduced in Section IV without a systematic study; this is a domain-specific adjustment that affects the number of clusters found.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Bi-cross validation for estimating spectral clustering hyper parameters." pith.science (2026). https://pith.science/paper/BRZO2KO6

@misc{pith2026190803747,
  author       = {Pith},
  title        = {Pith review of: Bi-cross validation for estimating spectral clustering hyper parameters},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/BRZO2KO6}},
  note         = {Machine review of arXiv:1908.03747}
}
read the original abstract

One challenge impeding the analysis of terabyte scale x-ray scattering data from the Linac Coherent Light Source LCLS, is determining the number of clusters required for the execution of traditional clustering algorithms. Here we demonstrate that previous work using bi-cross validation (BCV) to determine the number of singular vectors directly maps to the spectral clustering problem of estimating both the number of clusters and hyper parameter values. These results indicate that the process of estimating the number of clusters should not be divorced from the process of estimating other hyper parameters. Applying this method to LCLS x-ray scattering data enables the identification of dropped shots without manually setting boundaries on detector fluence and provides a path towards identifying rare and anomalous events.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

34 extracted references · 25 canonical work pages

  1. [1]

    Ishikawa , author H

    author author T. Ishikawa , author H. Aoyagi , author T. Asaka , author Y. Asano , author N. Azumi , author T. Bizen , author H. Ego , author K. Fukami , author T. Fukui , author Y. Furukawa , et al. ,\ @noop journal journal nature photonics \ volume 6 ,\ pages 540 ( year 2012 ) NoStop

    work page 2012

  2. [2]

    author author D. J. \ Higley , author A. H. \ Reid , author Z. Chen , author L. L. \ Guyader , author O. Hellwig , author A. A. \ Lutman , author T. Liu , author P. Shafer , author T. Chase , author G. L. \ Dakovski , et al. ,\ @noop journal journal arXiv preprint arXiv:1902.04611 \ ( year 2019 ) NoStop

    work page arXiv 1902

  3. [3]

    Yang , author X

    author author J. Yang , author X. Zhu , author T. J. \ Wolf , author Z. Li , author J. P. F. \ Nunes , author R. Coffee , author J. P. \ Cryan , author M. G \"u hr , author K. Hegazy , author T. F. \ Heinz , et al. ,\ @noop journal journal Science \ volume 361 ,\ pages 64 ( year 2018 ) NoStop

    work page 2018

  4. [4]

    Nogly , author T

    author author P. Nogly , author T. Weinert , author D. James , author S. Carbajo , author D. Ozerov , author A. Furrer , author D. Gashi , author V. Borin , author P. Skopintsev , author K. Jaeger , et al. ,\ @noop journal journal Science \ volume 361 ,\ pages eaat0094 ( year 2018 ) NoStop

    work page 2018

  5. [5]

    Hong , author H

    author author K. Hong , author H. Cho , author R. W. \ Schoenlein , author T. K. \ Kim , \ and\ author N. Huse ,\ @noop journal journal Accounts of chemical research \ volume 48 ,\ pages 2957 ( year 2015 ) NoStop

    work page 2015

  6. [6]

    Thayer , author D

    author author J. Thayer , author D. Damiani , author C. Ford , author I. Gaponenko , author W. Kroeger , author C. O'Grady , author J. Pines , author T. Tookey , author M. Weaver , \ and\ author A. Perazzo ,\ @noop journal journal Journal of Applied Crystallography \ volume 49 ,\ pages 1363 ( year 2016 ) NoStop

    work page 2016

  7. [7]

    Droste , author L

    author author S. Droste , author L. Shen , author V. E. \ White , author E. Diaz-Jacobo , author R. Coffee , author S. Zohar , author A. H. \ Reid , author F. Tavella , author M. P. \ Minitti , author J. J. \ Turner , et al. ,\ in\ @noop booktitle CLEO: Science and Innovations \ ( organization Optical Society of America ,\ year 2019 )\ pp.\ pages SF3I--7 NoStop

    work page 2019

  8. [8]

    author author J. C. \ Spence ,\ @noop journal journal Structural Dynamics \ volume 4 ,\ pages 044027 ( year 2017 ) NoStop

    work page 2017

Show all 34 references
  1. [9]

    Kupitz , author J

    author author C. Kupitz , author J. L. \ Olmos Jr , author M. Holl , author L. Tremblay , author K. Pande , author S. Pandey , author D. Oberth \"u r , author M. Hunter , author M. Liang , author A. Aquila , et al. ,\ @noop journal journal Structural Dynamics \ volume 4 ,\ pag...

    2017

  2. [10]

    Schoenlein , author S

    author author R. Schoenlein , author S. Boutet , author M. Minitti , \ and\ author A. Dunne ,\ @noop journal journal Applied Sciences \ volume 7 ,\ pages 850 ( year 2017 ) NoStop

    2017

  3. [11]

    Damiani , author M

    author author D. Damiani , author M. Dubrovin , author I. Gaponenko , author W. Kroeger , author T. Lane , author A. Mitra , author C. O'Grady , author A. Salnikov , author A. Sanchez-Gonzalez , author D. Schneider , et al. ,\ @noop journal journal Journal of Applied Crystallo...

    2016

  4. [12]

    author author C. H. \ Yoon , author P. Schwander , author C. Abergel , author I. Andersson , author J. Andreasson , author A. Aquila , author S. Bajt , author M. Barthelmess , author A. Barty , author M. J. \ Bogan , et al. ,\ @noop journal journal Optics express \ volume 19 ,...

    2011

  5. [13]

    Tibshirani , author G

    author author R. Tibshirani , author G. Walther , \ and\ author T. Hastie ,\ @noop journal journal Journal of the Royal Statistical Society: Series B (Statistical Methodology) \ volume 63 ,\ pages 411 ( year 2001 ) NoStop

    2001

  6. [14]

    author author C. A. \ Sugar \ and\ author G. M. \ James ,\ @noop journal journal Journal of the American Statistical Association \ volume 98 ,\ pages 750 ( year 2003 ) NoStop

    2003

  7. [15]

    Tibshirani \ and\ author G

    author author R. Tibshirani \ and\ author G. Walther ,\ @noop journal journal Journal of Computational and Graphical Statistics \ volume 14 ,\ pages 511 ( year 2005 ) NoStop

    2005

  8. [16]

    Von Luxburg et al

    author author U. Von Luxburg et al. ,\ @noop journal journal Foundations and Trends in Machine Learning \ volume 2 ,\ pages 235 ( year 2010 ) NoStop

    2010

  9. [17]

    Fujita , author D

    author author A. Fujita , author D. Y. \ Takahashi , \ and\ author A. G. \ Patriota ,\ @noop journal journal Computational Statistics & Data Analysis \ volume 73 ,\ pages 27 ( year 2014 ) NoStop

    2014

  10. [18]

    Fu \ and\ author P

    author author W. Fu \ and\ author P. O. \ Perry ,\ @noop journal journal arXiv preprint arXiv:1702.02658 \ ( year 2017 ) NoStop

    2017 arXiv

  11. [19]

    author author A. B. \ Owen , author P. O. \ Perry , et al. ,\ @noop journal journal The annals of applied statistics \ volume 3 ,\ pages 564 ( year 2009 ) NoStop

    2009

  12. [20]

    Von Luxburg ,\ @noop journal journal Statistics and computing \ volume 17 ,\ pages 395 ( year 2007 ) NoStop

    author author U. Von Luxburg ,\ @noop journal journal Statistics and computing \ volume 17 ,\ pages 395 ( year 2007 ) NoStop

    2007

  13. [21]

    \ Chung , author W.-C

    author author K.-M. \ Chung , author W.-C. \ Kao , author C.-L. \ Sun , author L.-L. \ Wang , \ and\ author C.-J. \ Lin ,\ @noop journal journal Neural computation \ volume 15 ,\ pages 2643 ( year 2003 ) NoStop

    2003

  14. [22]

    Mezzadri ,\ @noop journal journal arXiv preprint math-ph/0609050 \ ( year 2006 ) NoStop

    author author F. Mezzadri ,\ @noop journal journal arXiv preprint math-ph/0609050 \ ( year 2006 ) NoStop

    2006 arXiv

  15. [23]

    author author P. O. \ Perry ,\ @noop journal journal arXiv preprint arXiv:0909.3052 \ ( year 2009 ) NoStop

    2009 arXiv

  16. [24]

    Pedregosa , author G

    author author F. Pedregosa , author G. Varoquaux , author A. Gramfort , author V. Michel , author B. Thirion , author O. Grisel , author M. Blondel , author P. Prettenhofer , author R. Weiss , author V. Dubourg , author J. Vanderplas , author A. Passos , author D. Cournapeau ,...

    2011

  17. [25]

    author author T. E. \ Oliphant ,\ @noop title A guide to NumPy ,\ Vol. volume 1 \ ( publisher Trelgol Publishing USA ,\ year 2006 ) NoStop

    2006

  18. [26]

    author author T. E. \ Oliphant ,\ @noop journal journal Computing in Science & Engineering \ volume 9 ,\ pages 10 ( year 2007 ) NoStop

    2007

  19. [27]

    Van Der Walt , author S

    author author S. Van Der Walt , author S. C. \ Colbert , \ and\ author G. Varoquaux ,\ @noop journal journal Computing in Science & Engineering \ volume 13 ,\ pages 22 ( year 2011 ) NoStop

    2011

  20. [28]

    @noop title Demonstration of bi-cross validation for determining spectral clustering parameters and cluster number , \ howpublished https://github.com/sioan/bcv_spectral_clustering/blob/master/spectral_clustering.ipynb ,\ note accessed: 2019-05-7 NoStop

    2019

  21. [29]

    Zohar \ and\ author J

    author author S. Zohar \ and\ author J. J. \ Turner ,\ @noop journal journal Optics letters \ volume 44 ,\ pages 243 ( year 2019 ) NoStop

    2019

  22. [30]

    aapmrev4-2.bst 2018-12-27 (MD) hand-edited version of aapmrev4-1.bst

    FUNCTION id.bst "aapmrev4-2.bst 2018-12-27 (MD) hand-edited version of aapmrev4-1.bst" ENTRY address archive archivePrefix author bookaddress booktitle chapter collaboration doi edition editor eid eprint howpublished institution isbn issn journal key language month note number...

    2018

  23. [31]

    aipnauth4-2.bst 2018-12-27 (MD) hand-edited version of apsauth4-1.bst

    FUNCTION id.bst "aipnauth4-2.bst 2018-12-27 (MD) hand-edited version of apsauth4-1.bst" ENTRY address archive archivePrefix author bookaddress booktitle chapter collaboration doi edition editor eid eprint howpublished institution isbn issn journal key language month note numbe...

    2018

  24. [32]

    aipnum4-2.bst 2018-12-27 (MD) hand-edited version of apsrev4-1.bst

    FUNCTION id.bst "aipnum4-2.bst 2018-12-27 (MD) hand-edited version of apsrev4-1.bst" ENTRY address archive archivePrefix author bookaddress booktitle chapter collaboration doi edition editor eid eprint howpublished institution isbn issn journal key language month note number o...

    2018

  25. [33]

    apsrev4-2.bst 2018-12-27 (MD) hand-edited version of apsrev4-1.bst

    FUNCTION id.bst "apsrev4-2.bst 2018-12-27 (MD) hand-edited version of apsrev4-1.bst" ENTRY address archive archivePrefix author bookaddress booktitle chapter collaboration doi edition editor eid eprint howpublished institution isbn issn journal key language month note number o...

    2018

  26. [34]

    merlin.mbs apsrmp4-1.bst 2010-07-25 4.21a (PWD, AO, DPC) hacked

    FUNCTION id.bst "merlin.mbs apsrmp4-1.bst 2010-07-25 4.21a (PWD, AO, DPC) hacked" ENTRY address archive archivePrefix author bookaddress booktitle chapter collaboration doi edition editor eid eprint howpublished institution isbn issn journal key language month note number orga...

    2010

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.