NeurIPS 2026 Reviewing Guidelines
Reviewing Guidelines for Different Contribution Types
Authors should select the Contribution Type that best fits their submission.
- General: Most submissions will fall into this category.
- Theory: The main contribution is via theoretical analyses and proofs.
- Use-Inspired: The main contribution is in framing or designing approaches to meet the needs of a specific real-world application. (This often involves, e.g., engaging with domain experts.)
- Concept & Feasibility: The main contribution is a highly novel, high potential reward idea with scope beyond what can be validated in a single paper. (The significance and originality bar for these contributions is high.)
- Negative Results: The main contribution is in understanding a negative result. (The significance and originality bar for these contributions is high.)
The review form is the same for all Contribution Types, but the way that reviewing criteria should be interpreted differ across Types. Reviewers should assess a submission according to the Contribution Type selected by the authors. It is not possible for Contribution Type to be changed after submission, either by the authors or by the reviewers.
Below, we provide important guidance for Reviewers and ACs on how to interpret the overall reviewing criteria for each Contribution Type.
General Reviewing Guidelines
Quality
Is the submission technically sound? Are claims well supported (e.g., by theoretical analysis or experimental results)? Are the methods used appropriate? Is this a complete piece of work or work in progress? Are the authors careful and honest about evaluating both the strengths and weaknesses of their work?
Clarity
Is the submission clearly written? Is it well organized? Does it adequately inform the reader? A superbly written paper provides enough information for an expert reader to reproduce its results.
Significance
Are the results impactful for the community? Are others (researchers or practitioners) likely to use the ideas or build on them? Does the submission address a difficult task in a better way than previous work? Does it advance our understanding/knowledge on the topic in a demonstrable way? Does it provide unique data, unique conclusions about existing data, or a unique theoretical or experimental approach?
Originality
Does the work provide new insights, deepen understanding, or highlight important properties of existing methods? Is it clear how this work differs from previous contributions, with relevant citations provided? Does the work introduce novel tasks, problem framings, metrics, or methods that advance the field? Does this work offer a novel combination of existing techniques, and is the reasoning behind this combination well-articulated? As the questions above indicates, originality does not necessarily require introducing an entirely new method. Rather, a work that provides novel insights by evaluating existing methods, or demonstrates improved efficiency, fairness, etc. is also equally valuable.
Theory Reviewing Guidelines
Quality
Mathematical rigor and correctness. The primary criterion is the correctness of the claims. Proofs, lemmas, and the overall logical flow must be mathematically sound. While reviewers are not expected to verify every line, they should examine the core arguments sufficiently to establish a high level of confidence in the results.
Appropriateness of assumptions. Theoretical results may require assumptions, but the standard for "good" assumptions varies by the type of contribution. Appropriateness of assumptions should be weighed against the novelty of the result and standard norms of the existing literature.
Empirical validation is not necessary. Theoretical contributions may stand on their own and the purpose of designing new algorithms in this context need not be to outperform state-of-the-art applied models or methods. Empirical evaluation is not a necessary component, and a theory paper shouldn’t be penalized for lacking experiments. If empirical results are included, their function can be to further study formalized insights, not necessarily to compete with state-of-the-art applied models or the largest datasets.
Clarity
Intuition alongside rigor. Clarity requires both readability of the artifact and whether authors should outline the high-level proof strategy or the intuition behind a new mathematical definition before diving into technical details.
Clear scoping of the contribution. The paper must clearly signpost its type of theoretical contribution early on and the distinction between novel contributions and prior work must be clear.
Significance
Significance should be evaluated based on the specific type of theoretical contribution the paper aims to make. Reviewers may look for impact in one or more of the following areas:
Novel Abstractions and Formulations. Does the paper propose a new, theoretically rigorous mathematical framework or definition for studying interesting phenomena? Good abstractions provide the community with a new vocabulary to study emerging technologies and problems at a level appropriate for the phenomenon being studied.
Progress on Established Problems. Does the work make significant progress on a community-studied problem, surface a new angle of study on a well-established problem that has faced bottlenecks, or solve an open problem?
Originality
Technical novelty. Originality can take several forms depending on the paper's goal. It might be a fundamentally new proof technique, a novel synthesis of tools from other disciplines (e.g., statistical physics, pure mathematics), or a completely new way to parameterize or define a problem.
Use-Inspired Reviewing Guidelines
Quality
Is the use case real and meaningful? A use case should not be artificially constructed to present an interesting methodological problem. Rather, it should arise from the pre-existing needs of users outside the NeurIPS community, and this motivation should be made clear.
Is the design matched to the use case? Use-inspired submissions should closely tie their task framing, methods, metrics, etc. to the needs of a use case. For example, methods may need to accommodate the structure of the data, incorporate physical constraints, or be evaluated using metrics such as interpretability or robustness, depending on the use case.
Expect "non-standard" datasets. Submissions in this Contribution Type will frequently be evaluated on real-world datasets that fall outside common ML benchmarks. Such datasets should be encouraged if they are justified in relation to the use case.
Clarity
Clarity should be measured in relation to an ML audience. While the use case in question may provide motivation for the submission, the authors should not require the reader to have expertise in the domain of application and should explain any relevant jargon.
Significance
Significance can encompass both the potential for impact on an important use case and impact for the broader NeurIPS community. For this reason, novel ideas that are simple to apply in practice may be especially valuable.
Prior work outside the ML literature should be considered in addition to ML work. If the paper presents novel methods, for example, these should be compared against any commonly used non-ML approaches as well as baseline ML methods.
Originality
Originality need not mean wholly novel methods. It could mean a novel combination of existing methods to match the characteristics of the data, or a novel way of framing a stereotypical ML problem due to the particular needs of the use case.
Papers should provide insights to guide methods development. Thus, a paper may apply existing methods in a new context, but in such cases the discussion should analyze why certain methodological choices are effective or ineffective.
Concept and Feasibility Reviewing Guidelines
Quality
Strongly supported claims. As with papers in the general category, the submission must be technically sound and claims must be rigorously grounded through some combination of empirical, analytical, and conceptual arguments. For Concept and Feasibility papers, the scope of the ideas presented may be bigger than can be validated within a single paper, but must be strongly supported nonetheless. (This is in contrast to workshop submissions, where ideas may be only partially supported or in progress.)
Clarity
Is the submission clearly written? Is it well organized? Does it adequately inform the reader? A superbly written paper provides enough information for an expert reader to reproduce its results.
Significance
High potential for changing our approaches and/or understanding. The significance bar for Concept and Feasibility papers is high. The promise of a Concept and Feasibility Paper goes significantly beyond the specific evaluations included in the submission. Whether it changes how an important class of problems may be solved or significantly changes our understanding of machine learning, it is a paper that has the potential to change paradigms and/or practice in a substantive and significant way. The evaluations should be chosen to indicate to the reader this broader significance of the work.
Originality
Highly novel ideas. The originality bar for Concept and Feasibility papers is high. Submissions must contain highly novel ideas that change how problems or methods may be approached.
Negative Results Reviewing Guidelines
Quality
Grounded analysis and experimentation. In addition to the quality standards for a general paper, it is important that the negative result not be simply an empirical observation that some experiment did not turn out as expected or hoped. It is important that a negative result be grounded in deeper analysis, whether through a combination of conceptually-informed conjectures and careful experimentation, through rigorous proofs, or some combination.
Clarity
Is the submission clearly written? Is it well organized? Does it adequately inform the reader? A superbly written paper provides enough information for an expert reader to reproduce its results.
Significance
Changes our understanding of an important question. The significance bar for Negative Results papers is high. The negative result should be informative in changing how the community addresses a question or exposing that the community should be addressing an important question in a different way. A negative result relating to something few people would need or want is not significant.
A Negative Results paper need not identify a path to mitigate the negative finding; that can be left to future work.
Originality
Unexpected or surprising in some way. The originality bar for Negative Results papers is high. The negative result must be surprising or unexpected in some way - that is, it should run counter to a popularly held understanding. “A linear classifier cannot separate a nonlinear decision boundary” is a negative result that can be rigorously shown (quality) with clean language (clarity) and matters a lot (significance). However, it is not original because readers may be expected to either know this fact or be unsurprised to hear it.