Thoughtful research. Clearer manuscripts. More confident submissions.
Computer Science & Artificial Intelligence: Machine learning, data science, information systems, software engineering, and computational research.
Much of computer science, and machine learning and artificial intelligence in particular, publishes through peer-reviewed conferences with fixed deadlines, strict page limits, and venue-specific templates. In other areas, such as information systems, journals carry more weight. Knowing which convention applies shapes how long a paper can be, how it is reviewed, and how quickly it reaches readers.
Reviewers increasingly ask whether results can be reproduced. They look for clear descriptions of datasets, models, hyperparameters, and compute, and many venues ask for code and data availability statements or a reproducibility checklist. Combined with double-blind review and widespread preprint posting, these expectations make preparation a matter of procedural care as well as technical clarity.
In many subfields a full paper at a selective conference is a primary research output, not a preliminary one. Each call for papers sets page limits, formatting rules, supplementary material policies, and deadlines that are rarely extended, so planning backward from the deadline matters.
Many venues use double-blind review, which means removing names, acknowledgments, and grant numbers; citing your own earlier work in the third person; and sharing code or data through anonymized links. Venues differ on whether preprints, public repositories, and talks about the work are acceptable during review.
Reviewers expect enough detail to reproduce results: dataset versions and splits, preprocessing, architectures, hyperparameters, random seeds, number of runs, variation across runs, and compute resources. Some venues formalize this; NeurIPS, for example, requires authors to complete its paper checklist (external link), and some venues evaluate released code and data as artifacts.
Papers that introduce or depend on datasets should explain how the data were collected, labeled, filtered, and licensed, and what their limitations are. Structured documentation such as Datasheets for Datasets (external link) is widely used, and consent, privacy, and licensing questions apply to scraped or human-generated data.
Terms such as “model,” “agent,” “benchmark,” and “generalization” carry specific meanings in different subcommunities and should be defined once and used consistently, alongside a single notation for variables and functions. Policies on disclosing generative AI use in writing or research vary by venue and publisher, so check the current call for papers.
Fitting a contribution into a fixed page count often pushes essential details into appendices. Reviewers are not always required to read supplementary material, so the main text must contain enough to judge the method and results.
Self-identifying phrases, unredacted repository links, file metadata, and acknowledgments can reveal who wrote a paper. Some venues desk reject submissions that break anonymity, so a final check before upload is worth the time.
State-of-the-art claims based on a single run, a narrow benchmark, or untuned baselines draw skeptical reviews. Reporting variation across runs, ablations, and fair baseline comparisons makes the claims easier to accept.
Conference rebuttal periods are brief and often length-limited. Responses need to be prioritized, precise, and courteous, addressing the concerns most likely to change a reviewer’s assessment; responding to reviewer comments point by point covers the principles.
After a rejection, papers are often revised quickly for another venue with a different template, page limit, and reviewer pool. Carrying the reviewers’ substantive points forward while reformatting is easy to shortchange; see how to assess a rejection and plan your next step.
More on this in how to choose a journal beyond its impact factor.
Tightens prose to fit page limits while keeping technical terminology, notation, and claims consistent.
Adapts a paper to a new venue’s template, page limit, reference style, and anonymization rules when moving between conferences or to a journal.
Helps structure concise rebuttals within short author-response windows, and fuller response letters for journal revisions.
Supports clear abstracts, posters, and plain-language summaries for work presented at conferences and shared as preprints.
Official guidance worth reading alongside your target journal’s instructions.
A clear response letter answers every comment, shows exactly what changed and where, and handles disagreement with evidence. Here is a method that keeps the letter and manuscript in step.
Peer Review6 min read
Not every rejection means the same thing. Reading the decision letter closely tells you whether to revise, retarget, appeal, or resubmit, and what to fix first.
Peer Review7 min read
Editors, reviewers, and search tools often see only your title and abstract. Here is how to make both accurate, specific, and easy to find, without overselling the findings.
Research Communication6 min read
It depends on the venue. Many computer science conferences allow preprints but some restrict how the work is publicized during review, and reviewers are usually asked not to seek out authors’ identities. Check the current call for papers, and note that arXiv’s moderation (external link) is not peer review.
Typically the datasets and versions used, how to obtain the code, key hyperparameters and training details, the number of runs and how variation is reported, and the compute used. Checklists such as the NeurIPS paper checklist (external link) make these expectations explicit, and a justified “no” is generally better than an item left unexplained.
Policies differ across venues and publishers, and they continue to change. Some ask about any generative AI use beyond basic editing; others distinguish between writing assistance and use in the research method itself. Read the policy for your target venue, and remember that authors remain responsible for every claim and citation in the paper.
Computing spans very different communities, from theory and systems to human-computer interaction and information systems. Suitability for a particular paper is confirmed during consultation, once its topic, methods, venue, and deadline are clear. Share the details of your paper and venue to start.
Working on something specific?
Tell us the topic and the journal you have in mind. We confirm fit before proposing anything.
Ask about your manuscriptTell us the topic, the study design, and where you hope to submit. We’ll confirm whether we can help well before proposing anything.
