Polished and Well Written

A well-written paper is clear and easy to read. A polished paper is in nearly publication-ready form. The relevant question when reviewing is not whether the paper could become clear or well-written after a substantial rewrite, but whether the submitted paper is already clear and well-written. The authors are asking for your time and the time of others to review their work, and so it is reasonable for you to hold them to a high standard with regard to the quality of the work they submit.

Acceptance should be based on the paper as submitted. A paper should be accepted only if it is ready for publication in its current form, apart from a small number of minor edits. Typos and small presentation issues are inevitable, but they should be rare. If the paper needs a nontrivial editing pass, restructuring, notation cleanup, figure redesign, reference cleanup, or other substantial presentation work, then it should receive revisions at best. If the submission requires more significant changes, or if it appears that the authors did not put in an appropriate effort to polish their paper prior to submission, then rejection is appropriate.

Reviewers should focus on clarity, not elegance. Papers should not be penalized merely because the prose is plain, non-native, or stylistically different from what the reviewer would write. The standard is whether the intended RL audience can understand the problem, claims, methods, evidence, and limitations without unreasonable effort.

What Counts as Polished and Well Written?

A polished and well-written paper should generally have the following properties.

The paper should have a clear structure. The introduction should explain the problem setting and the paper's main claims. Sections and paragraphs should have clear purposes. Important transitions should not be left for the reader to infer.

The paper should be readable by RL researchers in general, not only by experts in a narrow subarea. Specialized material is often necessary, but the paper should provide enough context, definitions, and explanation for a broad RL audience to understand what is being claimed and how the claims are supported.

The paper should clearly state the problem setting when this is needed to understand the contribution. This includes the relevant assumptions, notation, objects of study, and evaluation criteria. Symbols, acronyms, and technical terms should be defined before they are used.

The paper should use notation consistently. The same object should not be referred to by multiple names or symbols without explanation. Different objects should not share the same notation unless the overloading is standard and clear from context.

The paper should clearly separate background and prior work from the authors' own contributions. This does not prescribe a particular section structure; rather, it should be clear to the reader which ideas, methods, and results are established and which are introduced by the authors.

Figures, tables, algorithms, theorem statements, and experimental results should be integrated into the text. They should be referenced and explained where they are used. A reader should not have to guess why a figure or table appears, what question it answers, or how it supports a claim. Mathematical expressions should be explained in English before they appear, so that readers do not have to figure out what the expression means from the math alone.

Figures and tables should be readable at normal viewing size. Axes, legends, labels, line styles, captions, and annotations should be legible. Captions (or text surrounding or prior to the reference) should explain what is shown, including the relevant experimental setting, metric, and meaning of error bars or shaded regions when applicable.

References should be complete, accurate, and consistently formatted. Citations should be used in a grammatically correct way and should make clear which prior work is being discussed. Poorly prepared references are a sign that the paper is not ready for publication.

The paper should follow RLJ formatting and style requirements. It should not use formatting tricks that harm readability or fairness, such as fonts that are too small, unreadable figure text, excessive compression of content, or violations of page or anonymity requirements.

The paper should not contain visible signs of being unfinished, such as unresolved TODOs, placeholder text, missing figures, broken cross-references, duplicated text, inconsistent section numbering, compilation artifacts, or unexplained notes to the authors. While some of these may occasionally slip through, multiple such errors are grounds for rejection (as opposed to revisions) if the authors have not put the appropriate effort into polishing their submission prior to demanding the time of their peers to evaluate their work.

Common Presentation Issues

Reviewers and authors may consult the "Avoiding Common Errors" section of the JMLR formatting instructions for examples of common style and formatting problems. RLJ does not require authors to copy JMLR's style exactly, but that section gives useful examples of the level of care expected in an archival journal article. For more on what constitutes using references in a grammatically correct way, see the "Citations" subsection of that same section.

Common issues include, but are not limited to:

• citations that are not grammatically integrated into sentences;
• mathematical expressions that are not punctuated as part of the surrounding sentence;
• incomplete or inconsistent references;
• figures with text that is too small to read;
• sections that begin immediately with a subsection without any orienting text;
• inconsistent capitalization of fields, methods, or technical terms;
• unclear (or inconsistent) hyphenation in compound modifiers;
• inconsistent or incorrect use of abbreviations;
• unnecessary formal wording that makes the paper harder to read without adding meaning.

A single minor issue of this kind should rarely affect a decision. Repeated issues of this kind are evidence that the paper is not polished.

Severity Calibration

Reviewers should distinguish minor presentation problems from problems that require revisions or rejection.

Minor issues: compatible with acceptance

Minor issues are small problems that can be fixed quickly and do not slow understanding. These can include a small number of typos, occasional awkward sentences, a few reference-formatting inconsistencies, an isolated notation slip where the intended meaning is clear, or a minor caption improvement.

Accepted papers do not need to be perfect, and authors of accepted papers will have a chance to make minor changes. However, the remaining issues should be few enough that the Editor does not need a response from the authors before publication.

Moderate issues: revisions at best

Moderate issues are problems that do not necessarily prevent evaluation of the paper, but do mean that the paper is not ready for publication as submitted. These would likely result in revisions. Examples include:

• many typos, grammar errors, or awkward sentences that slow reading;
• insufficient explanation of mathematical expressions that places an undue burden on the reader to interpret the expressions themselves;
• sections that are hard to follow because they lack signposting or transitions;
• undefined notation or terminology that appears repeatedly;
• figures or tables with missing labels, unclear captions, unreadable text, or unexplained quantities;
• a problem setting that is stated too late, too vaguely, or in a way that requires the reader to reconstruct it;
• repeated copyediting and formatting errors of the kind described in the JMLR guidance.

These problems may be fixable, but they require more than minor edits. A paper with these issues should not be accepted as-is.

Severe issues: rejection

A sufficient number of the previously described moderate issues can together form a severe issue. It is the authors' responsibility to carefully proofread their submissions prior to submission and to put in the time and effort required to craft a well-written and polished submission. Moderate issues (and a decision to recommend revisions) occur when authors make the appropriate effort to create a well-written and polished submission, but one issue (or a few issues) slip through. When there are enough moderate issues to indicate that authors did not invest the appropriate time and effort polishing their submission, these issues together form a severe issue.

In other cases, individual issues can be severe, and can result in rejection. Examples include:

• the paper is so poorly organized that the reviewer cannot reliably determine what is being claimed;
• the problem setting, notation, or terminology is too unclear to evaluate the technical content;
• the writing is so unclear that reviewers cannot separate presentation problems from technical problems;
• the paper contains unresolved TODOs, placeholders, missing figures, or other signs that it was not ready for review;
• formatting violations make the paper hard to read, hard to evaluate, or unfairly compressed.

What Reviewers Should Write

Reviewers are not expected to copyedit the paper. They should not list every typo or rewrite the paper for the authors. Instead, they should identify recurring problems and give representative examples.

A useful comment explains the type of issue, gives enough examples to establish its severity, and states whether the issue affects readability, evaluation, or publication readiness.

For example, instead of writing "the notation is unclear," a reviewer might write:

Several key symbols are used before being defined, including XXX in Section 2 and YYY in Algorithm 1. The notation for the learned policy also changes between Sections 3 and 4. These issues made it difficult to verify which quantities are being optimized.

Instead of writing "the writing needs proofreading," a reviewer might write:

The paper contains frequent grammar errors, broken cross-references, and several incomplete captions. These issues occur throughout the paper and slow down reading.

Reviewers should state the severity of the problem. Minor presentation issues may be noted briefly. Moderate issues should be described clearly enough for the Editor to understand why the paper is not ready for acceptance. Severe issues should explain why the presentation prevents reliable evaluation or would require a major rewrite.

Decision Rule

Accept only if the paper is already clear, coherent, readable, and professionally prepared. Recommend revisions when the paper is fundamentally understandable but needs a nontrivial editing or presentation pass before publication. Recommend rejection when the presentation problems are so extensive that the paper would need a major rewrite, cannot be reliably evaluated, or appears substantially unpolished.