Formatting Instructions For NeurIPS 2024

Anonymous Author(s)
Affiliation
Address
email

Abstract
1 The abstract paragraph should be indented 1/2 inch (3 picas) on both the left- and
2 right-hand margins. Use 10 point type, with a vertical spacing (leading) of 11 points.
3 The word Abstract must be centered, bold, and in point size 12. Two line spaces
4 precede the abstract. The abstract must be limited to one paragraph.

5 1 Submission of papers to NeurIPS 2024

6 Please read the instructions below carefully and follow them faithfully.

7 1.1 Style

8 Papers to be submitted to NeurIPS 2024 must be prepared according to the instructions presented
9 here. Papers may only be up to nine pages long, including figures. Additional pages containing only
10 acknowledgments and references are allowed. Papers that exceed the page limit will not be reviewed,
11 or in any other way considered for presentation at the conference.
12 The margins in 2024 are the same as those in previous years.
13 Authors are required to use the NeurIPS LATEX style files obtainable at the NeurIPS website as
14 indicated below. Please make sure you use the current files and not previous versions. Tweaking the
15 style files may be grounds for rejection.

16 1.2 Retrieval of style files

17 The style files for NeurIPS and other conference information are available on the website at
18 http://www.neurips.cc/
19 The file neurips_2024.pdf contains these instructions and illustrates the various formatting re-
20 quirements your NeurIPS paper must satisfy.
21 The only supported style file for NeurIPS 2024 is neurips_2024.sty, rewritten for LATEX 2ε .
22 Previous style files for LATEX 2.09, Microsoft Word, and RTF are no longer supported!
23 The LATEX style file contains three optional arguments: final, which creates a camera-ready copy,
24 preprint, which creates a preprint for submission to, e.g., arXiv, and nonatbib, which will not
25 load the natbib package for you in case of package clash.

26 Preprint option If you wish to post a preprint of your work online, e.g., on arXiv, using the
27 NeurIPS style, please use the preprint option. This will create a nonanonymized version of your
28 work with the text “Preprint. Work in progress.” in the footer. This version may be distributed as you
29 see fit, as long as you do not say which conference it was submitted to. Please do not use the final
30 option, which should only be used for papers accepted to NeurIPS.

Submitted to 38th Conference on Neural Information Processing Systems (NeurIPS 2024). Do not distribute.
31 At submission time, please omit the final and preprint options. This will anonymize your
32 submission and add line numbers to aid review. Please do not refer to these line numbers in your
33 paper as they will be removed during generation of camera-ready copies.
34 The file neurips_2024.tex may be used as a “shell” for writing your paper. All you have to do is
35 replace the author, title, abstract, and text of the paper with your own.
36 The formatting instructions contained in these style files are summarized in Sections 2, 3, and 4
37 below.

38 2 General formatting instructions

39 The text must be confined within a rectangle 5.5 inches (33 picas) wide and 9 inches (54 picas) long.
40 The left margin is 1.5 inch (9 picas). Use 10 point type with a vertical spacing (leading) of 11 points.
41 Times New Roman is the preferred typeface throughout, and will be selected for you by default.
42 Paragraphs are separated by 1/2 line space (5.5 points), with no indentation.
43 The paper title should be 17 point, initial caps/lower case, bold, centered between two horizontal
44 rules. The top rule should be 4 points thick and the bottom rule should be 1 point thick. Allow 1/4 inch
45 space above and below the title to rules. All pages should start at 1 inch (6 picas) from the top of the
46 page.
47 For the final version, authors’ names are set in boldface, and each name is centered above the
48 corresponding address. The lead author’s name is to be listed first (left-most), and the co-authors’
49 names (if different address) are set to follow. If there is only one co-author, list both author and
50 co-author side by side.
51 Please pay special attention to the instructions in Section 4 regarding figures, tables, acknowledgments,
52 and references.

53 3 Headings: first level

54 All headings should be lower case (except for first word and proper nouns), flush left, and bold.
55 First-level headings should be in 12-point type.

56 3.1 Headings: second level

57 Second-level headings should be in 10-point type.

58 3.1.1 Headings: third level

59 Third-level headings should be in 10-point type.

60 Paragraphs There is also a \paragraph command available, which sets the heading in bold, flush
61 left, and inline with the text, with the heading followed by 1 em of space.

62 4 Citations, figures, tables, references

63 These instructions apply to everyone.

64 4.1 Citations within the text

65 The natbib package will be loaded for you by default. Citations may be author/year or numeric, as
66 long as you maintain internal consistency. As to the format of the references themselves, any style is
67 acceptable as long as it is used consistently.
68 The documentation for natbib may be found at

69 http://mirrors.ctan.org/macros/latex/contrib/natbib/natnotes.pdf

2
 Figure 1: Sample figure caption.

70 Of note is the command \citet, which produces citations appropriate for use in inline text. For
71 example,

72 \citet{hasselmo} investigated\dots

73 produces

74 Hasselmo, et al. (1995) investigated. . .

75 If you wish to load the natbib package with options, you may add the following before loading the
76 neurips_2024 package:

77 \PassOptionsToPackage{options}{natbib}

78 If natbib clashes with another package you load, you can add the optional argument nonatbib
79 when loading the style file:

80 \usepackage[nonatbib]{neurips_2024}

81 As submission is double blind, refer to your own published work in the third person. That is, use “In
82 the previous work of Jones et al. [4],” not “In our previous work [4].” If you cite your other papers
83 that are not widely available (e.g., a journal paper under review), use anonymous author names in the
84 citation, e.g., an author of the form “A. Anonymous” and include a copy of the anonymized paper in
85 the supplementary material.

86 4.2 Footnotes

87 Footnotes should be used sparingly. If you do require a footnote, indicate footnotes with a number1
88 in the text. Place the footnotes at the bottom of the page on which they appear. Precede the footnote
89 with a horizontal rule of 2 inches (12 picas).
90 Note that footnotes are properly typeset after punctuation marks.2

91 4.3 Figures

92 All artwork must be neat, clean, and legible. Lines should be dark enough for purposes of reproduction.
93 The figure number and caption always appear after the figure. Place one line space before the figure
94 caption and one line space after the figure. The figure caption should be lower case (except for first
95 word and proper nouns); figures are numbered consecutively.
96 You may use color figures. However, it is best for the figure captions and the paper body to be legible
97 if the paper is printed in either black/white or in color.

1
Sample of the first footnote.
2
As in this example.

3
 Table 1: Sample table title

Part
Name Description Size (µm)
Dendrite Input terminal ∼100
Axon Output terminal ∼10
Soma Cell body up to 106

98 4.4 Tables

99 All tables must be centered, neat, clean and legible. The table number and title always appear before
100 the table. See Table 1.
101 Place one line space before the table title, one line space after the table title, and one line space after
102 the table. The table title must be lower case (except for first word and proper nouns); tables are
103 numbered consecutively.
104 Note that publication-quality tables do not contain vertical rules. We strongly suggest the use of the
105 booktabs package, which allows for typesetting high-quality, professional tables:

106 https://www.ctan.org/pkg/booktabs

107 This package was used to typeset Table 1.

108 4.5 Math

109 Note that display math in bare TeX commands will not create correct line numbers for sub-
110 mission. Please use LaTeX (or AMSTeX) commands for unnumbered display math. (You
111 really shouldn’t be using $$ anyway; see https://tex.stackexchange.com/questions/
112 503/why-is-preferable-to and https://tex.stackexchange.com/questions/40492/
113 what-are-the-differences-between-align-equation-and-displaymath for more infor-
114 mation.)

115 4.6 Final instructions

116 Do not change any aspects of the formatting parameters in the style files. In particular, do not modify
117 the width or length of the rectangle the text should fit into, and do not change font sizes (except
118 perhaps in the References section; see below). Please note that pages should be numbered.

119 5 Preparing PDF files

120 Please prepare submission files with paper size “US Letter,” and not, for example, “A4.”
121 Fonts were the main cause of problems in the past years. Your PDF file must only contain Type 1 or
122 Embedded TrueType fonts. Here are a few instructions to achieve this.

123 • You should directly generate PDF files using pdflatex.

124 • You can check which fonts a PDF files uses. In Acrobat Reader, select the menu
125 Files>Document Properties>Fonts and select Show All Fonts. You can also use the program
126 pdffonts which comes with xpdf and is available out-of-the-box on most Linux machines.
127 • xfig "patterned" shapes are implemented with bitmap fonts. Use "solid" shapes instead.
128 • The \bbold package almost always uses bitmap fonts. You should use the equivalent AMS
129 Fonts:
130 \usepackage{amsfonts}
131 followed by, e.g., \mathbb{R}, \mathbb{N}, or \mathbb{C} for R, N or C. You can also
132 use the following workaround for reals, natural and complex:

4
133 \newcommand{\RR}{I\!\!R} %real numbers
134 \newcommand{\Nat}{I\!\!N} %natural numbers
135 \newcommand{\CC}{I\!\!\!\!C} %complex numbers
136 Note that amsfonts is automatically loaded by the amssymb package.

137 If your file contains type 3 fonts or non embedded TrueType fonts, we will ask you to fix it.

138 5.1 Margins in LATEX

139 Most of the margin problems come from figures positioned by hand using \special or other
140 commands. We suggest using the command \includegraphics from the graphicx package.
141 Always specify the figure width as a multiple of the line width as in the example below:

142 \usepackage[pdftex]{graphicx} ...

143 \includegraphics[width=0.8\linewidth]{myfile.pdf}

144 See Section 4.4 in the graphics bundle documentation (http://mirrors.ctan.org/macros/

145 latex/required/graphics/grfguide.pdf)
146 A number of width problems arise when LATEX cannot properly hyphenate a line. Please give LaTeX
147 hyphenation hints using the \- command when necessary.

148 References
149 References follow the acknowledgments in the camera-ready paper. Use unnumbered first-level
150 heading for the references. Any choice of citation style is acceptable as long as you are consistent. It
151 is permissible to reduce the font size to small (9 point) when listing the references. Note that the
152 Reference section does not count towards the page limit.

153 [1] Alexander, J.A. & Mozer, M.C. (1995) Template-based algorithms for connectionist rule extraction. In
154 G. Tesauro, D.S. Touretzky and T.K. Leen (eds.), Advances in Neural Information Processing Systems 7, pp.
155 609–616. Cambridge, MA: MIT Press.
156 [2] Bower, J.M. & Beeman, D. (1995) The Book of GENESIS: Exploring Realistic Neural Models with the
157 GEneral NEural SImulation System. New York: TELOS/Springer–Verlag.
158 [3] Hasselmo, M.E., Schnell, E. & Barkai, E. (1995) Dynamics of learning and recall at excitatory recurrent
159 synapses and cholinergic modulation in rat hippocampal region CA3. Journal of Neuroscience 15(7):5249-5262.

160 A Appendix / supplemental material

161 Optionally include supplemental material (complete proofs, additional experiments and plots) in
162 appendix. All such materials SHOULD be included in the main submission.

5
163 NeurIPS Paper Checklist
164 The checklist is designed to encourage best practices for responsible machine learning research,
165 addressing issues of reproducibility, transparency, research ethics, and societal impact. Do not remove
166 the checklist: The papers not including the checklist will be desk rejected. The checklist should
167 follow the references and precede the (optional) supplemental material. The checklist does NOT
168 count towards the page limit.
169 Please read the checklist guidelines carefully for information on how to answer these questions. For
170 each question in the checklist:

171 • You should answer [Yes] , [No] , or [NA] .

172 • [NA] means either that the question is Not Applicable for that particular paper or the
173 relevant information is Not Available.
174 • Please provide a short (1–2 sentence) justification right after your answer (even for NA).

175 The checklist answers are an integral part of your paper submission. They are visible to the
176 reviewers, area chairs, senior area chairs, and ethics reviewers. You will be asked to also include it
177 (after eventual revisions) with the final version of your paper, and its final version will be published
178 with the paper.
179 The reviewers of your paper will be asked to use the checklist as one of the factors in their evaluation.
180 While "[Yes] " is generally preferable to "[No] ", it is perfectly acceptable to answer "[No] " provided a
181 proper justification is given (e.g., "error bars are not reported because it would be too computationally
182 expensive" or "we were unable to find the license for the dataset we used"). In general, answering
183 "[No] " or "[NA] " is not grounds for rejection. While the questions are phrased in a binary way, we
184 acknowledge that the true answer is often more nuanced, so please just use your best judgment and
185 write a justification to elaborate. All supporting evidence can appear either in the main paper or the
186 supplemental material, provided in appendix. If you answer [Yes] to a question, in the justification
187 please point to the section(s) where related material for the question can be found.
188 IMPORTANT, please:

189 • Delete this instruction block, but keep the section heading “NeurIPS paper checklist",
190 • Keep the checklist subsection headings, questions/answers and guidelines below.
191 • Do not modify the questions and only use the provided macros for your answers.

192 1. Claims
193 Question: Do the main claims made in the abstract and introduction accurately reflect the
194 paper’s contributions and scope?
195 Answer: [TODO]
196 Justification: [TODO]
197 Guidelines:
198 • The answer NA means that the abstract and introduction do not include the claims
199 made in the paper.
200 • The abstract and/or introduction should clearly state the claims made, including the
201 contributions made in the paper and important assumptions and limitations. A No or
202 NA answer to this question will not be perceived well by the reviewers.
203 • The claims made should match theoretical and experimental results, and reflect how
204 much the results can be expected to generalize to other settings.
205 • It is fine to include aspirational goals as motivation as long as it is clear that these goals
206 are not attained by the paper.
207 2. Limitations
208 Question: Does the paper discuss the limitations of the work performed by the authors?
209 Answer: [TODO]
210 Justification: [TODO]

6
211 Guidelines:
212 • The answer NA means that the paper has no limitation while the answer No means that
213 the paper has limitations, but those are not discussed in the paper.
214 • The authors are encouraged to create a separate "Limitations" section in their paper.
215 • The paper should point out any strong assumptions and how robust the results are to
216 violations of these assumptions (e.g., independence assumptions, noiseless settings,
217 model well-specification, asymptotic approximations only holding locally). The authors
218 should reflect on how these assumptions might be violated in practice and what the
219 implications would be.
220 • The authors should reflect on the scope of the claims made, e.g., if the approach was
221 only tested on a few datasets or with a few runs. In general, empirical results often
222 depend on implicit assumptions, which should be articulated.
223 • The authors should reflect on the factors that influence the performance of the approach.
224 For example, a facial recognition algorithm may perform poorly when image resolution
225 is low or images are taken in low lighting. Or a speech-to-text system might not be
226 used reliably to provide closed captions for online lectures because it fails to handle
227 technical jargon.
228 • The authors should discuss the computational efficiency of the proposed algorithms
229 and how they scale with dataset size.
230 • If applicable, the authors should discuss possible limitations of their approach to
231 address problems of privacy and fairness.
232 • While the authors might fear that complete honesty about limitations might be used by
233 reviewers as grounds for rejection, a worse outcome might be that reviewers discover
234 limitations that aren’t acknowledged in the paper. The authors should use their best
235 judgment and recognize that individual actions in favor of transparency play an impor-
236 tant role in developing norms that preserve the integrity of the community. Reviewers
237 will be specifically instructed to not penalize honesty concerning limitations.
238 3. Theory Assumptions and Proofs
239 Question: For each theoretical result, does the paper provide the full set of assumptions and
240 a complete (and correct) proof?
241 Answer: [TODO]
242 Justification: [TODO]
243 Guidelines:
244 • The answer NA means that the paper does not include theoretical results.
245 • All the theorems, formulas, and proofs in the paper should be numbered and cross-
246 referenced.
247 • All assumptions should be clearly stated or referenced in the statement of any theorems.
248 • The proofs can either appear in the main paper or the supplemental material, but if
249 they appear in the supplemental material, the authors are encouraged to provide a short
250 proof sketch to provide intuition.
251 • Inversely, any informal proof provided in the core of the paper should be complemented
252 by formal proofs provided in appendix or supplemental material.
253 • Theorems and Lemmas that the proof relies upon should be properly referenced.
254 4. Experimental Result Reproducibility
255 Question: Does the paper fully disclose all the information needed to reproduce the main ex-
256 perimental results of the paper to the extent that it affects the main claims and/or conclusions
257 of the paper (regardless of whether the code and data are provided or not)?
258 Answer: [TODO]
259 Justification: [TODO]
260 Guidelines:
261 • The answer NA means that the paper does not include experiments.

7
262 • If the paper includes experiments, a No answer to this question will not be perceived
263 well by the reviewers: Making the paper reproducible is important, regardless of
264 whether the code and data are provided or not.
265 • If the contribution is a dataset and/or model, the authors should describe the steps taken
266 to make their results reproducible or verifiable.
267 • Depending on the contribution, reproducibility can be accomplished in various ways.
268 For example, if the contribution is a novel architecture, describing the architecture fully
269 might suffice, or if the contribution is a specific model and empirical evaluation, it may
270 be necessary to either make it possible for others to replicate the model with the same
271 dataset, or provide access to the model. In general. releasing code and data is often
272 one good way to accomplish this, but reproducibility can also be provided via detailed
273 instructions for how to replicate the results, access to a hosted model (e.g., in the case
274 of a large language model), releasing of a model checkpoint, or other means that are
275 appropriate to the research performed.
276 • While NeurIPS does not require releasing code, the conference does require all submis-
277 sions to provide some reasonable avenue for reproducibility, which may depend on the
278 nature of the contribution. For example
279 (a) If the contribution is primarily a new algorithm, the paper should make it clear how
280 to reproduce that algorithm.
281 (b) If the contribution is primarily a new model architecture, the paper should describe
282 the architecture clearly and fully.
283 (c) If the contribution is a new model (e.g., a large language model), then there should
284 either be a way to access this model for reproducing the results or a way to reproduce
285 the model (e.g., with an open-source dataset or instructions for how to construct
286 the dataset).
287 (d) We recognize that reproducibility may be tricky in some cases, in which case
288 authors are welcome to describe the particular way they provide for reproducibility.
289 In the case of closed-source models, it may be that access to the model is limited in
290 some way (e.g., to registered users), but it should be possible for other researchers
291 to have some path to reproducing or verifying the results.
292 5. Open access to data and code
293 Question: Does the paper provide open access to the data and code, with sufficient instruc-
294 tions to faithfully reproduce the main experimental results, as described in supplemental
295 material?
296 Answer: [TODO]
297 Justification: [TODO]
298 Guidelines:
299 • The answer NA means that paper does not include experiments requiring code.
300 • Please see the NeurIPS code and data submission guidelines (https://nips.cc/
301 public/guides/CodeSubmissionPolicy) for more details.
302 • While we encourage the release of code and data, we understand that this might not be
303 possible, so “No” is an acceptable answer. Papers cannot be rejected simply for not
304 including code, unless this is central to the contribution (e.g., for a new open-source
305 benchmark).
306 • The instructions should contain the exact command and environment needed to run to
307 reproduce the results. See the NeurIPS code and data submission guidelines (https:
308 //nips.cc/public/guides/CodeSubmissionPolicy) for more details.
309 • The authors should provide instructions on data access and preparation, including how
310 to access the raw data, preprocessed data, intermediate data, and generated data, etc.
311 • The authors should provide scripts to reproduce all experimental results for the new
312 proposed method and baselines. If only a subset of experiments are reproducible, they
313 should state which ones are omitted from the script and why.
314 • At submission time, to preserve anonymity, the authors should release anonymized
315 versions (if applicable).

8
316 • Providing as much information as possible in supplemental material (appended to the
317 paper) is recommended, but including URLs to data and code is permitted.
318 6. Experimental Setting/Details
319 Question: Does the paper specify all the training and test details (e.g., data splits, hyper-
320 parameters, how they were chosen, type of optimizer, etc.) necessary to understand the
321 results?
322 Answer: [TODO]
323 Justification: [TODO]
324 Guidelines:
325 • The answer NA means that the paper does not include experiments.
326 • The experimental setting should be presented in the core of the paper to a level of detail
327 that is necessary to appreciate the results and make sense of them.
328 • The full details can be provided either with the code, in appendix, or as supplemental
329 material.
330 7. Experiment Statistical Significance
331 Question: Does the paper report error bars suitably and correctly defined or other appropriate
332 information about the statistical significance of the experiments?
333 Answer: [TODO]
334 Justification: [TODO]
335 Guidelines:
336 • The answer NA means that the paper does not include experiments.
337 • The authors should answer "Yes" if the results are accompanied by error bars, confi-
338 dence intervals, or statistical significance tests, at least for the experiments that support
339 the main claims of the paper.
340 • The factors of variability that the error bars are capturing should be clearly stated (for
341 example, train/test split, initialization, random drawing of some parameter, or overall
342 run with given experimental conditions).
343 • The method for calculating the error bars should be explained (closed form formula,
344 call to a library function, bootstrap, etc.)
345 • The assumptions made should be given (e.g., Normally distributed errors).
346 • It should be clear whether the error bar is the standard deviation or the standard error
347 of the mean.
348 • It is OK to report 1-sigma error bars, but one should state it. The authors should
349 preferably report a 2-sigma error bar than state that they have a 96% CI, if the hypothesis
350 of Normality of errors is not verified.
351 • For asymmetric distributions, the authors should be careful not to show in tables or
352 figures symmetric error bars that would yield results that are out of range (e.g. negative
353 error rates).
354 • If error bars are reported in tables or plots, The authors should explain in the text how
355 they were calculated and reference the corresponding figures or tables in the text.
356 8. Experiments Compute Resources
357 Question: For each experiment, does the paper provide sufficient information on the com-
358 puter resources (type of compute workers, memory, time of execution) needed to reproduce
359 the experiments?
360 Answer: [TODO]
361 Justification: [TODO]
362 Guidelines:
363 • The answer NA means that the paper does not include experiments.
364 • The paper should indicate the type of compute workers CPU or GPU, internal cluster,
365 or cloud provider, including relevant memory and storage.

9
366 • The paper should provide the amount of compute required for each of the individual
367 experimental runs as well as estimate the total compute.
368 • The paper should disclose whether the full research project required more compute
369 than the experiments reported in the paper (e.g., preliminary or failed experiments that
370 didn’t make it into the paper).
371 9. Code Of Ethics
372 Question: Does the research conducted in the paper conform, in every respect, with the
373 NeurIPS Code of Ethics https://neurips.cc/public/EthicsGuidelines?
374 Answer: [TODO]
375 Justification: [TODO]
376 Guidelines:
377 • The answer NA means that the authors have not reviewed the NeurIPS Code of Ethics.
378 • If the authors answer No, they should explain the special circumstances that require a
379 deviation from the Code of Ethics.
380 • The authors should make sure to preserve anonymity (e.g., if there is a special consid-
381 eration due to laws or regulations in their jurisdiction).
382 10. Broader Impacts
383 Question: Does the paper discuss both potential positive societal impacts and negative
384 societal impacts of the work performed?
385 Answer: [TODO]
386 Justification: [TODO]
387 Guidelines:
388 • The answer NA means that there is no societal impact of the work performed.
389 • If the authors answer NA or No, they should explain why their work has no societal
390 impact or why the paper does not address societal impact.
391 • Examples of negative societal impacts include potential malicious or unintended uses
392 (e.g., disinformation, generating fake profiles, surveillance), fairness considerations
393 (e.g., deployment of technologies that could make decisions that unfairly impact specific
394 groups), privacy considerations, and security considerations.
395 • The conference expects that many papers will be foundational research and not tied
396 to particular applications, let alone deployments. However, if there is a direct path to
397 any negative applications, the authors should point it out. For example, it is legitimate
398 to point out that an improvement in the quality of generative models could be used to
399 generate deepfakes for disinformation. On the other hand, it is not needed to point out
400 that a generic algorithm for optimizing neural networks could enable people to train
401 models that generate Deepfakes faster.
402 • The authors should consider possible harms that could arise when the technology is
403 being used as intended and functioning correctly, harms that could arise when the
404 technology is being used as intended but gives incorrect results, and harms following
405 from (intentional or unintentional) misuse of the technology.
406 • If there are negative societal impacts, the authors could also discuss possible mitigation
407 strategies (e.g., gated release of models, providing defenses in addition to attacks,
408 mechanisms for monitoring misuse, mechanisms to monitor how a system learns from
409 feedback over time, improving the efficiency and accessibility of ML).
410 11. Safeguards
411 Question: Does the paper describe safeguards that have been put in place for responsible
412 release of data or models that have a high risk for misuse (e.g., pretrained language models,
413 image generators, or scraped datasets)?
414 Answer: [TODO]
415 Justification: [TODO]
416 Guidelines:
417 • The answer NA means that the paper poses no such risks.

10
418 • Released models that have a high risk for misuse or dual-use should be released with
419 necessary safeguards to allow for controlled use of the model, for example by requiring
420 that users adhere to usage guidelines or restrictions to access the model or implementing
421 safety filters.
422 • Datasets that have been scraped from the Internet could pose safety risks. The authors
423 should describe how they avoided releasing unsafe images.
424 • We recognize that providing effective safeguards is challenging, and many papers do
425 not require this, but we encourage authors to take this into account and make a best
426 faith effort.
427 12. Licenses for existing assets
428 Question: Are the creators or original owners of assets (e.g., code, data, models), used in
429 the paper, properly credited and are the license and terms of use explicitly mentioned and
430 properly respected?
431 Answer: [TODO]
432 Justification: [TODO]
433 Guidelines:
434 • The answer NA means that the paper does not use existing assets.
435 • The authors should cite the original paper that produced the code package or dataset.
436 • The authors should state which version of the asset is used and, if possible, include a
437 URL.
438 • The name of the license (e.g., CC-BY 4.0) should be included for each asset.
439 • For scraped data from a particular source (e.g., website), the copyright and terms of
440 service of that source should be provided.
441 • If assets are released, the license, copyright information, and terms of use in the
442 package should be provided. For popular datasets, paperswithcode.com/datasets
443 has curated licenses for some datasets. Their licensing guide can help determine the
444 license of a dataset.
445 • For existing datasets that are re-packaged, both the original license and the license of
446 the derived asset (if it has changed) should be provided.
447 • If this information is not available online, the authors are encouraged to reach out to
448 the asset’s creators.
449 13. New Assets
450 Question: Are new assets introduced in the paper well documented and is the documentation
451 provided alongside the assets?
452 Answer: [TODO]
453 Justification: [TODO]
454 Guidelines:
455 • The answer NA means that the paper does not release new assets.
456 • Researchers should communicate the details of the dataset/code/model as part of their
457 submissions via structured templates. This includes details about training, license,
458 limitations, etc.
459 • The paper should discuss whether and how consent was obtained from people whose
460 asset is used.
461 • At submission time, remember to anonymize your assets (if applicable). You can either
462 create an anonymized URL or include an anonymized zip file.
463 14. Crowdsourcing and Research with Human Subjects
464 Question: For crowdsourcing experiments and research with human subjects, does the paper
465 include the full text of instructions given to participants and screenshots, if applicable, as
466 well as details about compensation (if any)?
467 Answer: [TODO]
468 Justification: [TODO]

11
469 Guidelines:
470 • The answer NA means that the paper does not involve crowdsourcing nor research with
471 human subjects.
472 • Including this information in the supplemental material is fine, but if the main contribu-
473 tion of the paper involves human subjects, then as much detail as possible should be
474 included in the main paper.
475 • According to the NeurIPS Code of Ethics, workers involved in data collection, curation,
476 or other labor should be paid at least the minimum wage in the country of the data
477 collector.
478 15. Institutional Review Board (IRB) Approvals or Equivalent for Research with Human
479 Subjects
480 Question: Does the paper describe potential risks incurred by study participants, whether
481 such risks were disclosed to the subjects, and whether Institutional Review Board (IRB)
482 approvals (or an equivalent approval/review based on the requirements of your country or
483 institution) were obtained?
484 Answer: [TODO]
485 Justification: [TODO]
486 Guidelines:
487 • The answer NA means that the paper does not involve crowdsourcing nor research with
488 human subjects.
489 • Depending on the country in which research is conducted, IRB approval (or equivalent)
490 may be required for any human subjects research. If you obtained IRB approval, you
491 should clearly state this in the paper.
492 • We recognize that the procedures for this may vary significantly between institutions
493 and locations, and we expect authors to adhere to the NeurIPS Code of Ethics and the
494 guidelines for their institution.
495 • For initial submissions, do not include any information that would break anonymity (if
496 applicable), such as the institution conducting the review.

12