-
Opening day
-
Submission deadline
The Fast Machine Learning for Science Conference is an all-plenary conference. Tutorials and topical sessions are parallel.
We welcome abstract submissions for the following presentation formats:
- Presentations and/or Posters: Posters will be shown during 1.5-hour poster sessions. The committee will select the 15-minute spotlight presentations. New for 2026: Submitters will also have the option to submit a 4-page extended abstract (paper) through OpenReview (details to follow).
- Tutorials: In-depth tutorials (1.5-3 hours) on Monday, August 31. Specify level (beginner/intermediate/advanced).
- Birds-of-a-Feather (BoF) Sessions: Topical, participant-driven discussions (1.5-3 hours) on Thursday, September 3 or Friday, September 4.
Topics
Topics include, but are not limited to
Machine Learning Algorithm Design & Optimization
- Novel efficient architectures
- Hyperparameter optimization and tuning
- Model compression (quantization, sparsity)
- Hardware/software co-design for ML efficiency
- Large language model optimization and implementation
Accelerated Inference & Real-Time Processing
- Low-latency ML for scientific experiments
- FPGA/NPU/GPU-based ML acceleration
- ML for trigger systems and data acquisition
- On-detector and edge inference
Scientific Applications of Fast ML
- High-energy physics, astrophysics, and astronomy
- Space science and satellite-based ML
- Genomics and medical imaging
- Climate and environmental modeling
- Biological science and neuroscience
- Fusion
- Quantum computing
- Material science
- Robotics
Scalable & Distributed ML Systems
- Cloud-based, accelerated ML processing
- Distributed inference
- Acceleration-as-a-service
- ML compilers and runtimes
- Benchmarks and datasets
Advanced Hardware & Computing Architectures
- Specialized AI accelerators
- Tools and methodologies for accelerating ML algorithms
- Heterogeneous computing platforms for ML
- Beyond CMOS
Extended Abstracts
We will also accept 4-page extended abstracts to be submitted and reviewed through OpenReview. Accepted contributions will be displayed on OpenReview and Indico.
Instructions for Preparing Submissions
Authors may submit a 4-page short paper (a.k.a extended abstract) in PDF format. Submissions should use the official template specified on the submission page. References are unlimited and do not count toward the 4-page limit. Appendices are discouraged.
Papers should be submitted through the conference OpenReview. The organizers reserve the right to desk reject submissions that do not follow the formatting requirements, exceed the page limit, or otherwise fail to satisfy the submission instructions.
Submission Policy
Submissions do not need to be anonymized. Authors may include names, affiliations, acknowledgments, links to code or data repositories, and references to their own prior work where appropriate.
We welcome original work, work in progress, and extended-abstract versions of relevant work that has appeared or is under review elsewhere, provided that the submission fits the scope of Fast Machine Learning for Science. Authors should clearly indicate when the submission is based on previously published or publicly available work.
The conference is intended to encourage discussion and exchange rather than serve as an archival venue. Authors remain responsible for ensuring that their submission complies with the policies of any other venue where related work has been submitted or published.
Submission Requirements
Submissions should satisfy the following requirements:
- Maximum length: 4 pages, excluding references.
- Format: PDF using the NeurIPS template.
- References: unlimited and excluded from the page limit.
- Appendices: allowed but discouraged.
- Author list: all authors must be included at the time of submission.
- Confidentiality: submissions will remain confidential during review.
Accepted submissions will be made public on the FastML26 indico website.
A paper checklist or broader impact statement is not required. Authors may include a short broader impact statement if they wish, and it will not count toward the 4-page limit.
Review Process
Submissions that satisfy the formatting and policy requirements will be sent for peer review through OpenReview. There will be no rebuttal period. To improve review consistency and help reviewers provide clear feedback, the organizers may use large language models (LLMs) as assistive tools during the review process. Any LLM-assisted review workflow will be conducted with attention to confidentiality. Submissions and reviews should not be uploaded to public or personal LLM services unless explicitly approved by the organizers and consistent with the confidentiality requirements of the review process.
Review Criteria
Reviewers will be asked to consider the following criteria:
- Relevance to fast machine learning for science;
- Technical correctness and scientific soundness;
- Clarity and accessibility to a broad interdisciplinary audience;
- Novelty or usefulness of the method, result, dataset, benchmark, tool, or perspective;
- Potential impact on scientific discovery, real-time processing, accelerated inference, scalable ML systems, or efficient ML deployment;
- Quality of evidence supporting the claims;
- Value of the contribution to the conference community.
Poster Instructions
Posters should be no larger than 35" (0.89 m) wide by 48" (1.2 m) tall, with a portrait orientation. Posters can remain up throughout the event, though it is recommended that they don't remain up overnight.
Local printing service
FedEx
8849 Villa La Jolla Drive
La Jolla, CA 92037
Important Deadlines
- Abstract Submission:
June 1, 2026June 15, 2026 - (Optional) Extended Abstract Submission:
June 15, 2026July 1, 2026 - (Optional) Extended Abstract Camera-Ready Deadline: August 15, 2026