betaFeedback

WorksComputer Science

Truly Every Author is First Author: Equal-Width Superimposition by Minimising Visual Distortion

Abstract

In an earlier work, Demaine and Demaine propose that every author be a first author by superimposing all author names on the same spot of the title page. The construction is typographically unstable, because names have different natural widths. The superimposed block is ragged and the longest name dominates the visual mass, reintroducing exactly the ordering advantage the standard removes. We close this loophole. Each name is distorted by two knobs, a horizontal glyph scale and an inter-letter tracking, until all names occupy one common width WW. We give a quadratic distortion model, solve the per-name trade-off in closed form by Lagrange multipliers, and show that the total-distortion optimal common width is a stiffness-weighted mean of the natural widths. We provide implementations as a LaTeX\LaTeX package and a Typst package, covering author blocks and bibliography entries alike; this paper typesets its own author block and its own reference with the former. Every proposition is machine-checked in Lean. The LaTeX\LaTeX and Typst packages, as well as the proofs, are available at https://github.com/MrP01/EveryAuthorIsFirstAuthor.

Introduction

The convention that author order encodes credit is a lossy encoding of a partial order into a total one. Erik D. Demaine Martin L. Demaine [2] resolve it by making the encoding degenerate, superimposing all author names at the same position, so that no name precedes another.

AMOR [4] takes the opposite route and keeps the list, but shuffles it at random whenever it is viewed, so that no order is fixed. This removes the ranking across views but not within any single one, where some name still comes first; superimposition removes it in every view, and is the construction we build on.

Superimposition removes the horizontal ordering but not the horizontal extent. If, for example, Yu Liu and Martin L. Demaine are overprinted, the shorter name is contained in the longer one, and the reader’s eye recovers a ranking from ink coverage, where the widest name frames the block and the narrowest hides inside it. Formally, superimposition quotients out translation but leaves width as a visible, and thus creditable, per-author quantity. In Tufte’s terms [1], the block has an unbounded lie factor: it displays a width difference where the data, equal credit, has none.

We therefore require that all names be typeset to one common width. Since names cannot be padded with whitespace - padding is invisible under superimposition and merely restores the containment problem - the width must be achieved by distorting the letters themselves.

Two distortions are available, and they differ in character. Scaling the glyphs horizontally changes letterform, while tracking changes only the gaps. Both are ugly in excess. The engineering question is how to divide a required width change between them, and what common width to aim at in the first place.

Sections 2 and 3 answer both in closed form. Section 4 describes the implementations and a machine-checked proof of every claim, and Section 5 the results.

Problem Statement

Fix a font and a size. All lengths are in printer’s points. The model is stated for a single name and carries no index; the index appears only where names are compared.

Definition 1 (Name data). For a name we write

  • g∈N≥0g \in\mathbb{N}_{\ge0}, the number of inter-character gaps, that is, the number of positions at which tracking is inserted;

  • c=g+1c = g + 1, the number of characters;

  • w>0w > 0, the natural set width, as the font sets it;

  • e=w/ce = w/c, the mean character width, which is the natural length scale of the name.

Section 4 obtains ww and gg by measurement, without inspecting the characters of the name.

Definition 2 (Distortion variables). A name is typeset with a horizontal glyph scale s>0s > 0, under which every character is drawn ss times as wide at unchanged height, and a tracking t∈Rt \in\mathbb{R}, an extra kern of length tt inserted into each of the gg gaps. The undistorted setting is (s,t)=(1,0)(s,t) = (1,0).

Because the scale acts on characters and the tracking on gaps, the two contributions are additive and the realised width of the distorted name is exactly

ω(s,t)=sw+gt.(1)\omega(s,t) = sw + gt. \tag*{(1)}

The symbol ω\omega is reserved for the width a name actually attains, and WW below for the common width all names are asked to attain.

Definition 3 (Distortion cost). With weights α,β>0\alpha,\beta> 0, the ugliness of a name is

C(s,t)=α2(s−1)2+β2(te)2.(2)C(s,t) = \frac{\alpha}{2}(s-1)^2 + \frac{\beta}{2}\left(\frac{t}{e}\right)^2. \tag*{(2)}

Both terms are quadratic and vanish at the undistorted setting, so small distortions are cheap and the cost grows without bound. The scale term is dimensionless by construction. The tracking term is made dimensionless by measuring tt in units of the mean character width ee: a kern of one average letter is equally offensive in a large font and a small one, and equally offensive in a name of five letters and one of twenty. The ratio β/α\beta/\alpha is the only free parameter of the model and expresses the designer’s relative tolerance: β/α→∞\beta/\alpha\to\infty forbids tracking and recovers pure scaling, β/α→0\beta/\alpha\to0 forbids scaling and recovers pure tracking.

Multiple Names

A document has n≥1n \ge1 names, indexed by i∈{1,…,n}i \in\{1,\ldots,n\}, and from here on the quantities of name ii carry that index, so that gig_i, cic_i, wiw_i, eie_i, sis_i, tit_i, ωi\omega_i and CiC_i mean for name ii what gg, cc, ww, ee, ss, tt, ω\omega and CC mean for a single name, and the same applies to quantities introduced later, such as the stiffness KK of § 3.1.

Definition 4 (Equal-width superimposition problem). Given {(wi,gi,ei)}i=1n\{(w_i,g_i,e_i)\}_{i=1}^{n} and α,β>0\alpha,\beta>0, find a common width W>0W>0 and per-name distortions (si,ti)(s_i,t_i) solving

min⁡W,(si,ti)∑i=1nCi(si,ti)\min_{W,(s_i,t_i)} \sum_{i=1}^{n} C_i(s_i,t_i)

s.t. ωi(si,ti)=W,i=1,…,n\omega_i(s_i,t_i)=W,\quad i=1,\ldots,n.

Section 3 solves (3) in two stages, the inner problem for fixed WW, then the outer problem over WW.

Closed-Form Solution

The problem separates. For a fixed common width the per-name trade-off has an explicit minimiser (§ 3.1), whose optimal value is quadratic in that width, so the outer choice of width is an explicit weighted mean (§ 3.2).

The Optimal Split for a Fixed Width

This subsection concerns one name at a time, so it needs no index. Fix the target width WW, and write x=s−1x=s-1 for the scale excess and d=W−wd=W-w for the width deficit; the constraint ω(s,t)=W\omega(s,t)=W becomes wx+gt=dwx+gt=d, while the cost (2) reads α2x2+β2(t/e)2\frac{\alpha}{2}x^2+\frac{\beta}{2}(t/e)^2 in these variables.

Definition 5 (Stiffness). The stiffness of a name is

K=w2α+g2e2β>0.(3)K=\frac{w^2}{\alpha}+\frac{g^2e^2}{\beta}>0. \tag*{(3)}

KK measures how cheaply a name absorbs a width change. It is large when the name is long, since a given absolute change is then a small relative change, and small for short names. This is why KK appears as an inverse weight below.

Proposition 6 (Per-name split). For fixed WW, the unique minimiser of (2) subject to ω(s,t)=W\omega(s,t)=W is, with λ=(W−w)/K\lambda=(W-w)/K,

s=1+λwα,t=λge2β,(4)s=1+\frac{\lambda w}{\alpha},\qquad t=\frac{\lambda g e^2}{\beta}, \tag*{(4)}

and the optimal value is

C∗(W)=(W−w)22K.(5)C^*(W)=\frac{(W-w)^2}{2K}. \tag*{(5)}

Proof. The objective is strictly convex and the constraint affine, so the Karush-Kuhn-Tucker conditions are necessary and sufficient. With Lagrangian

L=α2x2+β2t2e2−λ(wx+gt−d),\mathcal{L}=\frac{\alpha}{2}x^2+\frac{\beta}{2}\frac{t^2}{e^2}-\lambda(wx+gt-d),

stationarity gives αx=λw\alpha x=\lambda w and βt/e2=λg\beta t/e^2=\lambda g, that is, x=λw/αx=\lambda w/\alpha and t=λge2/βt=\lambda g e^2/\beta. Substituting into the constraint,

d=wx+gt=λ(w2α+g2e2β)=λK,d=wx+gt=\lambda\left(\frac{w^2}{\alpha}+\frac{g^2e^2}{\beta}\right)=\lambda K,

so λ=d/K\lambda=d/K, which is (4). The value follows from α2x2+β2(t/e)2=λ2(wx+gt)=λd2=d2/(2K)\frac{\alpha}{2}x^2+\frac{\beta}{2}(t/e)^2=\frac{\lambda}{2}(wx+gt)=\frac{\lambda d}{2}=d^2/(2K). □

Two consequences are worth stating. First, the two distortions are used in the fixed ratio

gt(s−1)w=αg2e2βw2≈αβ,(6)\frac{gt}{(s-1)w}=\frac{\alpha g^2e^2}{\beta w^2}\approx\frac{\alpha}{\beta}, \tag*{(6)}

where the approximation uses ge≈wge\approx w and is exact up to one character. The width change is therefore split between tracking and scaling in essentially the ratio α:β\alpha:\beta, independently of WW, as Figure 1 shows; the weights are directly interpretable, and α=β=1\alpha=\beta=1 means “half the correction each”. Second, C∗C^* is quadratic in WW, which makes the outer problem trivial.

Figure 1. The Cost of Splitting a Fixed Width Deficit. The share ϕ\phi of the deficit taken by scaling costs αϕ2+β(1−ϕ)2\alpha\phi^2+\beta(1-\phi)^2 in units of d2/2w2d^2/2w^2, and the minimum (dot) sits at ϕ∗=β/(α+β)\phi^*=\beta/(\alpha+\beta), so the weights set the split, up to the approximation ge≈wge\approx w of the split ratio (6).

The Optimal Common Width

Proposition 7 (Optimal width). The minimiser of F(W)=∑iCi∗(W)F(W)=\sum_i C_i^*(W) is the stiffness-weighted mean of the natural widths,

W∗=∑i=1nwi/Ki∑i=1n1/Ki,(7)W^*=\frac{\sum_{i=1}^{n}w_i/K_i}{\sum_{i=1}^{n}1/K_i}, \tag*{(7)}

and it is unique.

Proof. By (5), F(W)=12∑i(W−wi)2/KiF(W)=\frac{1}{2}\sum_i(W-w_i)^2/K_i is a strictly convex quadratic in WW with F′′(W)=∑iKi−1>0F''(W)=\sum_i K_i^{-1}>0. Setting F′(W)=∑i(W−wi)/Ki=0F'(W)=\sum_i(W-w_i)/K_i=0 and solving for WW gives the weighted mean (7). □

Corollary 8. min⁡iwi≤W∗≤max⁡iwi\min_i w_i\le W^*\le\max_i w_i, with equality only if all wiw_i coincide.

So the optimal target is neither the widest name, which would stretch everyone else, nor the arithmetic mean width, but a mean that gives most weight to the names that are most expensive to move. Figure 2 shows the mechanism. Since KiK_i grows roughly like wi2w_i^2, long names are stiff, contribute little pull, and W∗W^* is biased towards the short names – correctly, because a short name pays the largest relative distortion for a given absolute change.

Figure 2. Choosing the Common Width for Five Names. Each name contributes a parabola Ci∗C_i^* whose vertex sits at its natural width wiw_i (grey dots) and whose curvature is 1/Ki1/K_i, so stiff names pull weakly. The total FF is minimised at W∗W^* (red dot), nearer the short names.

Feasibility and Clamping

Unbounded sis_i and tit_i are typographically unacceptable, so we impose si∈[smin⁡,smax⁡]s_i \in[s_{\min},s_{\max}], with defaults 0.85 and 1.15, and, implicitly, ti≥−eit_i \ge-e_i, at which tracking consumes the advance of an average letter. Letters touch well before that bound, which § 5 returns to. The constrained problem is still convex with a box constraint, and its solution is the projection of the unconstrained one followed by re-solving the equality constraint for the remaining free variable:

s^i=clamp⁡(si,smin⁡,smax⁡),t^i=W∗−s^iwigi(8)\hat{s}_i=\operatorname{clamp}(s_i,s_{\min},s_{\max}),\qquad\hat{t}_i=\frac{W^*-\hat{s}_i w_i}{g_i} \tag*{(8)}

for gi≥1g_i \ge1. This projection is exact. Along the width constraint tit_i is determined by sis_i, and the cost becomes a convex quadratic in sis_i alone, centred at the unconstrained optimum, so clamping sis_i to the box minimises it. Single-character names (gi=0g_i=0) have no gaps to track and must take si=W∗/wis_i=W^*/w_i regardless of the bounds; the model degenerates gracefully and stays feasible.

Implementation

LaTeX

The package equalauthors.sty implements (3) to (8) in expl3 floating point. Names are collected with \equalauthorsAddAuthor, and \equalauthorsAuthors measures each name and computes W∗W^* by (7) and the split by (4) and (8). Each name is then set as \scalebox{s}[1]{\textls[l]{name}} with s=sis=s_i and l=ℓi=1000ti/(si em)l=\ell_i=1000t_i/(s_i\,\mathrm{em}), the tracking in microtype’s units1, where the division by sis_i compensates for \scalebox scaling the inserted kerns too, and the boxes are overlapped in one box of width W∗W^*.

Splitting a name into characters would break UTF-8 input on 8-bit engines, strip accents written as commands, such as \"o, of their argument, and lose kerning pairs, so the package never takes a name apart and recovers the two quantities the model needs by measurement,

wi=ωi(1,0),gi=∂wi∂ti=ωi(1,em)−ωi(1,0)em.(9)w_i=\omega_i(1,0),\qquad g_i=\frac{\partial w_i}{\partial t_i}=\frac{\omega_i(1,\mathrm{em})-\omega_i(1,0)}{\mathrm{em}}. \tag*{(9)}

The second measurement sets the name with a tracking of one em, and since (1) is affine in tit_i, two \hbox_set:Nn measurements per name suffice. Microtype also places half the tracking outside the first and last letter, which would be invisible padding, so the package cancels it with a negative kern at either end. The remaining quotient equals the gap count, except that microtype stretches an interword space by 4/3 of the tracking rather than counting it as two gaps, so gig_i comes out fractional for names that contain spaces. Nothing in § 3 uses integrality, so the solution applies unchanged. The companion command \equalauthorsAuthorsStacked prints the names one per line at the common width (Figure 3).

Stacked names at the common width

Figure 3. Stacked Names at the Common Width. Left, the six names of Table 1, distorted and printed separately; right, the same set with Yu Liu in place of the long name. All lines of a block have the common width W⋆W^\star of that block.

Microtype rounds tracking to whole thousandths of an em, so the package re-solves the scale from si(wi+giτi)=W∗s_i(w_i+g_i\tau_i)=W^*, with τi\tau_i the quantised tracking handed to microtype. The width thus stays exactly W∗W^* while sis_i moves by about 10−310^{-3}, and at a bound ℓi\ell_i is rounded towards the inside. Weights and bounds are set with

\equalauthorsSetup{alpha=1, beta=1, smin=0.85,
  smax=1.15}

Citations

A reference list ranks authors exactly as a title page does, so the standard is only self-consistent if citations are superimposed as well. The command \equalauthorsList takes a BibTeX-style name list A and B and C, splits it on and, and applies (7) and (4) to the result. The macro is placed in the author field of the .bib entry inside an extra pair of braces,

author = {{\equalauthorsList{Erik D. Demaine and
  Martin L. Demaine}}}

which hides the and separators from biber and makes it pass the field through verbatim. Every reference of this paper is typeset this way, each with a common width W∗W^* over its own names alone.

Typst

The Typst package provides the same functions, using measure for wiw_i, scale for sis_i and the tracking parameter of text for tit_i. Grapheme clusters are a native notion there, so the encoding difficulties of § 4.1 do not arise and gig_i is counted directly. Typst, too, applies tracking inside the scaled coordinate frame, so the tracking handed to the engine is ti/sit_i/s_i; the realised width is then siwi+gitis_iw_i+g_it_i as in (1). A self-check compiles the distorted names and asserts that every measured width equals W∗W^* to within 10−410^{-4} pt.

Bibliographies need a different hook there. Typst renders references through a CSL style and exposes no element for the author list of an entry, so there is nothing to attach a show rule to. The hook is therefore moved into the citation style itself: equalauthors.csl is an ordinary numeric style whose only peculiarity is that it fences each entry’s author list between two invisible U+2062 characters, and the accompanying equal-references show rule matches exactly those runs and replaces them with the superimposed list.

Formal Verification

Every mathematical claim of Sections 2 to 4 and Appendix A is machine-checked in Lean 4 [3] against Mathlib [5], in lean/EqualAuthors.lean of the accompanying repository2, with every proof complete. This covers both propositions with uniqueness, the corollary, their composition into the full problem (3), the clamp (8), the measurement and rounding identities of the implementations, and the containment claim of the introduction; each theorem is named after the label it proves, such as prop_split. The proofs need no Lagrange multipliers, since along the constraint the cost exceeds its optimum by α2(s−s⋆)2+β2((t−t⋆)/e)2\frac{\alpha}{2}(s-s^\star)^2+\frac{\beta}{2}((t-t^\star)/e)^2, and F(W)F(W) exceeds F(W⋆)F(W^\star) by 12∑iKi−1(W−W⋆)2\frac{1}{2}\sum_i K_i^{-1}(W-W^\star)^2.

Results

The author block of this paper is an equal-width superimposition of its three authors. Table 1 reports the solution for a six-name example that adds the originators of the standard and one long name, and Figure 3 unstacks the same six names so that the common width can be checked by eye.

Name iiwiw_i [pt]sis_itit_i [pt]
Peter Waldert61.581.1550.173
Benedikt Kantz68.751.05180.063
Thomas Köhler67.861.06290.085
Erik D. Demaine74.060.9887-0.01
Martin L. Demaine84.190.8914-0.134
Jesús S. M. A.142.250.6056-0.466

Table 1. Natural Widths and Optimal Distortions at W⋆=73.09W^\star= 73.09 pt. The weights are α=1\alpha= 1 and β=4\beta= 4, and si∈[0.5,1.5]s_i \in[0.5, 1.5]. Names shorter than W⋆W^\star receive si>1s_i > 1 and ti>0t_i > 0; longer names receive both below their neutral values.

The long name, Jesús Salvador Martínez Alcaide, is on average twice as wide as the other five, yet at the settings of Table 1 it moves W⋆W^\star only from 69.8 to 73.1 pt, where the arithmetic mean would move by 12 pt. Its stiffness grows like wi2w_i^2, so it pulls with about a quarter of the weight of each of the others.

The same name shows where the default bounds end. At α=β=1\alpha=\beta=1 and si∈[0.85,1.15]s_i\in[0.85,1.15], its unconstrained scale falls below smin⁡s_{\min}, so the clamp (8) is active and tracking absorbs the remaining 48 pt, about −0.35ei-0.35e_i per gap. That is far above the bound −ei-e_i, yet adjacent letters already run into one another and the name becomes illegible. Table 1 therefore widens the scale range to [0.5,1.5][0.5,1.5] and penalises tracking four times as heavily, β=4\beta=4, which moves most of every correction onto the glyphs. The long name is then set at si≈0.61s_i\approx0.61 with ti≈−0.1eit_i\approx-0.1e_i, condensed but legible, and no clamp is active.

A very short name stresses the opposite end. Put in place of the long name, Yu Liu, at wi=30.1w_i=30.1 pt, is the least stiff name of the set and pulls W⋆W^\star from 69.8 down to 48.8 pt, as the right block of Figure 3 shows at the same settings. The other five names are condensed to between si≈0.65s_i\approx0.65 and 0.83, while Yu Liu reaches the bound smax⁡=1.5s_{\max}=1.5 and tracking adds the remainder, about 0.15ei0.15e_i per gap. Loose tracking is the milder distortion, so the short name stays legible, although it now sets everyone else’s width.

Limitations

The cost (2) models ugliness; it does not measure it. It is symmetric in the sign of the distortion, whereas readers tolerate loose tracking better than tight tracking, and condensed glyphs better than expanded ones; asymmetric weights α+,β+\alpha^+,\beta^+ would leave the convexity argument unchanged. Vertical extent is likewise creditable and left untouched, since names with ascenders and descenders still claim more ink.

We also skip opacity, which the original construction uses. Martin L. Demaine [2] set every name in semi-transparent ink, at opacity 2/3, so that the superimposed names do not merge into a solid black block, and overlapping strokes appear darker, which keeps the individual names legible. We print all names at full opacity. Combining the two is promising future work, in particular an opacity that decreases with the number of authors, as Martin L. Demaine suggests, since at fixed opacity the block grows darker with every name that is added.

Finally, the standard equalises typography, not contribution; nothing here prevents a single coauthor from still doing all the work.

Conclusion

Superimposing author names removes order but not width. Requiring equal width turns the layout into a small convex program whose solution is fully explicit. The split between glyph scaling and tracking is the ratio α:β\alpha:\beta of the two ugliness weights (Proposition 6), and the common width is the stiffness-weighted mean (7) of the natural widths (Proposition 7). Both are one line of arithmetic per name, so every author can be a first author of exactly the same size at no measurable cost in compile time.

Ablation Studies

Scaling Only

Changing only the glyph scale ss, that is, setting t=0t = 0 for every name, simplifies the model at the cost of ugliness. The realised width of a name reduces to ω(s)=sw\omega(s) = sw and its cost (2) to its first term,

C(s)=α2(s−1)2,C(s) = \frac{\alpha}{2}(s - 1)^2,

by which the Lagrangian simplifies to

L=α2x2−λ(wx−d).\mathcal{L} = \frac{\alpha}{2}x^2 - \lambda(wx - d).

Stationarity gives x=λw/αx = \lambda w/\alpha, which, substituted into the constraint, yields d=wx=λw2/αd = wx = \lambda w^2/\alpha and the optimal cost

C⋆(W)=α2x2=(W−w)22w2/α,C^\star(W) = \frac{\alpha}{2}x^2 = \frac{(W-w)^2}{2w^2/\alpha},

so K=w2/αK = w^2/\alpha, which is (3) without its tracking term. The common width then follows from (7) as in the general case, and every name takes s=W⋆/ws = W^\star/w. The bounds on ss cannot be enforced here, since no other variable is left to absorb the remainder. Figure 4 shows the result. Each name reaches W⋆W^\star through its glyph widths alone, so letterforms differ visibly between the lines, from s≈1.19s \approx1.19 for the shortest name to s≈0.51s \approx0.51 for the longest.

The six names with scaling only

Figure 4. The Six Names of Table 1 With Scaling Only, t=0t = 0. All six lines have width W⋆W^\star.

Tracking Only

Changing only the tracking tt, that is, setting s=1s = 1 for every name, simplifies the model in the same way and is uglier still. The realised width of a name reduces to ω(t)=w+gt\omega(t) = w + gt and its cost (2) to its second term,

C(t)=β2(te)2,C(t) = \frac{\beta}{2}\left(\frac{t}{e}\right)^2,

by which the Lagrangian simplifies to

L=βt22e2−λ(gt−d).\mathcal{L} = \frac{\beta t^2}{2e^2} - \lambda(gt - d).

Stationarity gives βt/e2=λg\beta t/e^2 = \lambda g, which, substituted into the constraint, yields d=gt=λg2e2/βd = gt = \lambda g^2e^2/\beta and the optimal cost

C⋆(W)=β2(te)2=(W−w)22g2e2/β,C^\star(W) = \frac{\beta}{2}\left(\frac{t}{e}\right)^2 = \frac{(W-w)^2}{2g^2e^2/\beta},

so K=g2e2/βK = g^2e^2/\beta, which is (3) without its scaling term. The common width is, again, computed by (7). Figure 5 shows that tracking alone cannot absorb large width changes. The long name needs about −0.5e-0.5e per gap and collapses, while the shortest name is visibly letterspaced.

The six names with tracking only

Figure 5. The Six Names of Table 1 With Tracking Only, s=1s = 1. All six lines have width W⋆W^\star.

References

  1. [1]Edward R. Tufte. The Visual Display of Quantitative Information. Cheshire, CT: Graphics Press, 1983.
  2. [2]Martin L. Demaine. Every Author as First Author. 2023. DOI: 10.48550/arXiv.2304.01393. arXiv: 2304.01393 [cs.DL].DOI
  3. [3]Leonardo de Moura and Sebastian Ullrich. “The Lean 4 Theorem Prover and Programming Language”. In: Automated Deduction – CADE 28. Vol. 12699. Lecture Notes in Computer Science. Springer, 2021, pp. 625–635. DOI: 10.1007/978-3-030-79876-5_37.DOI
  4. [4]Martin L. Demaine. AMOR: Ambiguous Authorship Order. 2024. DOI: 10.48550/arXiv.2404.00994. arXiv: 2404.00994 [cs.CV].
  5. [5]The mathlib Community. “The Lean Mathematical Library”. In: Proceedings of the 9th ACM SIGPLAN International Conference on Certified Programs and Proofs. ACM, 2020, pp. 367–381. DOI: 10.1145/3372885.3373824.DOI

Paper details

vlnc:W_q5qdfnvp
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Computer Science

Not recorded for this paper

View statement
Looking for related papers…
Contents