Written and maintained by CASRAI Editorial Board
Last updated
The NeurIPS acceptance rate is the percentage of submitted research papers accepted for presentation and publication in the proceedings of the Annual Conference on Neural Information Processing Systems. As one of the flagship peer-reviewed venues in machine learning and computational neuroscience, NeurIPS serves as a primary benchmark for selective research dissemination. Tracking its acceptance rate reveals broader trends in AI publication volume, reviewer workload scaling, and the competitive threshold for scientific contributions.
In modern computer science publishing, top-tier conference proceedings function as archival journal equivalents, carrying significant weight for academic hiring, tenure, and fellowship evaluation. Understanding how NeurIPS calculates its acceptance statistics—and how its selectivity compares to sister venues such as ICML and ICLR—helps research teams plan submission timelines, benchmark review distributions, and navigate the double-blind peer-review lifecycle.
Historical Acceptance Rates at NeurIPS (2020–2024)
Over the past five years, NeurIPS has experienced an unprecedented surge in submission volume, driven by advances in foundation models, generative architectures, and empirical machine learning. Despite the rapid growth in submissions, the conference has maintained a remarkably consistent acceptance rate hovering between 25% and 26% for its main conference track.
Starting in 2021, NeurIPS introduced the dedicated Datasets and Benchmarks Track to provide peer-reviewed venue status for empirical resources, data artifacts, and evaluation suites. The table below presents the official figures documented in the NeurIPS Fact Sheets published by the Neural Information Processing Systems Foundation.
| Conference Edition | Main Track Submissions | Main Track Accepted | Main Track Rate | Datasets Track Submissions | Datasets Track Accepted | Datasets Track Rate |
|---|---|---|---|---|---|---|
| NeurIPS 2024 (Vancouver) | 15,671 | 4,037 | 25.8% | 1,820 | 460 | 25.3% |
| NeurIPS 2023 (New Orleans) | 12,343 | 3,218 | 26.1% | 987 | 322 | 32.6% |
| NeurIPS 2022 (New Orleans) | 9,634 | 2,465 | 25.6% | 487 | 177 | 36.3% |
| NeurIPS 2021 (Virtual) | 9,122 | 2,344 | 25.7% | 490 | 174 | 35.5% |
| NeurIPS 2020 (Virtual) | 9,454 | 1,900 | 20.1% | — | — | — |
A notable trend is the maturation of the Datasets and Benchmarks track. While its inaugural editions in 2021–2023 exhibited acceptance rates between 32% and 36%, by 2024 submission volume grew by 84% (from 987 to 1,820 submissions), bringing the track’s acceptance rate down to 25.3%, closely converging with the main track rate.
NeurIPS vs. Other Top AI Conferences
To evaluate publication strategy, researchers frequently compare NeurIPS with peer venues in machine learning and artificial intelligence, including ICML (International Conference on Machine Learning), ICLR (International Conference on Learning Representations), AAAI (Association for the Advancement of Artificial Intelligence), AISTATS (International Conference on Artificial Intelligence and Statistics), and IJCAI (International Joint Conference on Artificial Intelligence).
| Conference Venue | Review Platform | 2024 Submissions | 2024 Accepted Papers | 2024 Acceptance Rate | Primary Review Model |
|---|---|---|---|---|---|
| NeurIPS 2024 | OpenReview | 15,671 (main) | 4,037 | 25.8% | Double-blind, public after decision |
| ICML 2024 | OpenReview | 9,473 | 2,609 | 27.5% | Double-blind, author rebuttal |
| ICLR 2024 | OpenReview | 7,262 | 2,260 | 31.0% | Double-blind, fully open discussion |
| AAAI 2024 | OpenReview / CMT | 9,862 | 2,342 | 23.8% | Double-blind, two-phase review |
| AISTATS 2024 | OpenReview | 1,980 | 547 | 27.6% | Double-blind, focused scope |
| IJCAI 2024 | OpenReview | 5,651 | 791 (main) | 14.0% | Double-blind, two-phase filtering |
While ICLR maintains a slightly higher acceptance yield (~31.0%), its review workflow features public discussion threads during the review period. NeurIPS and ICML historically maintain selective yields near 26% to 28%, balancing empirical depth with theoretical rigor. In contrast, IJCAI implements stringent early summary rejects that compress its main-track acceptance rate to approximately 14.0%.
Review Process Logistics: Area Chairs, Reviewer Scaling, and Ethics Checks
Maintaining review quality across more than 15,000 submissions requires massive administrative and technical infrastructure. The NeurIPS peer-review process relies on a hierarchical structure managed by Program Chairs, Senior Area Chairs (SACs), Area Chairs (ACs), and peer reviewers.
Official data from the NeurIPS 2024 Fact Sheet illustrates the scale of this operation:
- Reviewer Pool: 13,640 main conference reviewers and 1,844 dataset track reviewers evaluated submissions.
- Area Chair Workload: 1,393 main track ACs and 195 SACs managed meta-reviews, averaging approximately 11 to 12 papers per Area Chair stack.
- Ethics Review Workflow: 377 dedicated ethics reviewers evaluated 958 flagged papers (5.48% of all submissions in 2024, up from 502 papers or 3.77% in 2023). Submissions flagged for societal harm, dataset consent issues, or dual-use risk undergo this secondary evaluation before final acceptance.
Understanding this operational scale underscores why author rebuttals must remain concise and targeted: Area Chairs must synthesize multiple divergent reviews across dozens of manuscripts under tight deadlines, as detailed in our guide on writing a conference rebuttal.
Interpreting Scores, Rebuttals, and Decision Tiers
NeurIPS uses numerical scoring guidelines on OpenReview, where reviewers score manuscripts along soundness, presentation, and contribution scales alongside an overall recommendation ranging from 1 (strong reject) to 10 (award quality). Following initial reviews, authors enter a dedicated discussion period to respond to reviewer questions, clarify theoretical bounds, or present supplementary validation experiments.
Because acceptance decisions reflect holistic evaluation by Area Chairs rather than a simple arithmetic average, borderline papers (averaging around 5.5 to 6.0) often hinge on champion reviewers advocating for a novel insight. Accepted papers are stratified into distinct presentation tiers:
- Poster Presentations: Comprise the vast majority of accepted papers (typically over 85% of accepted manuscripts).
- Spotlight Presentations: Reserved for highly ranked papers (approximately 3% to 4% of total submissions) that receive abbreviated plenary or specialized session highlights.
- Oral Presentations: The most selective tier (under 1.5% of total submissions), representing top-scoring contributions selected for full-length podium talks.
For research teams balancing academic credit models across publishing formats, see our analysis of computer science conference vs. journal authorship credit and the operational mechanics of open peer review models.
Frequently Asked Questions
How does the Datasets and Benchmarks track acceptance rate compare to the main conference?
In its initial years (2021–2023), the NeurIPS Datasets and Benchmarks track maintained a higher acceptance rate between 32% and 36%. However, by NeurIPS 2024, submission volume doubled to 1,820 papers, and its acceptance rate normalized to 25.3%, aligning closely with the main track’s 25.8% rate.
How does the NeurIPS acceptance rate compare to ICML and ICLR?
Across recent editions, NeurIPS typically records an acceptance rate of 25% to 26%, ICML ranges between 27% and 28%, and ICLR averages around 31%. All three venues employ double-blind review managed through OpenReview, but differ in discussion openness and timeline structures.
What percentage of submitted papers get an oral or spotlight presentation at NeurIPS?
Less than 5% of total submissions receive oral or spotlight recognition. At NeurIPS 2024, oral presentations accounted for approximately 1.5% of submitted manuscripts, while spotlight presentations represented roughly 3.5%, with the remainder presented as conference posters.
Does paper withdrawal affect the final reported acceptance rate?
Official fact sheets calculate the acceptance rate against the net pool of evaluated submissions after desk rejections and early withdrawals. Papers withdrawn before reviewer assignment are excluded from the denominator, whereas papers withdrawn after receiving initial reviews remain tracked in institutional submission logs.
Sources
- Neural Information Processing Systems Foundation. (2024). NeurIPS 2024 Fact Sheet: 38th Annual Conference on Neural Information Processing Systems. Vancouver, BC, Canada. URL: https://media.neurips.cc/Conferences/NeurIPS2024/NeurIPS2024-Fact_Sheet.pdf (Read: 2026-11-07).
- Neural Information Processing Systems Foundation. (2023). NeurIPS 2023 Fact Sheet: 37th Annual Conference on Neural Information Processing Systems. New Orleans, LA, USA. URL: https://media.neurips.cc/Conferences/NeurIPS2023/NeurIPS2023-Fact_Sheet.pdf (Read: 2026-11-07).
- International Conference on Learning Representations. (2024). ICLR 2024 Fact Sheet: Twelfth Annual International Conference on Learning Representations. Vienna, Austria. URL: https://media.iclr.cc/Conferences/ICLR2024/ICLR2024-Fact_Sheet.pdf (Read: 2026-11-07).
- Proceedings of Machine Learning Research. (2024). Proceedings of the 41st International Conference on Machine Learning (ICML 2024). PMLR Volume 235. URL: https://proceedings.mlr.press/v235/ (Read: 2026-11-07).
- Association for the Advancement of Artificial Intelligence. (2024). The 38th Annual AAAI Conference on Artificial Intelligence (AAAI-24) Conference Statistics. Vancouver, BC, Canada. URL: https://aaai.org/conference/aaai/aaai-24/ (Read: 2026-11-07).
