Extended Abstracts
Up to 4-page submissions (not including references or appendices) describing novel research, systems, datasets, simulators, benchmarks, and methodologies in ML for Systems.
Machine Learning/CORE Unranked
MLForSys2026 2026
Machine Learning for Systems 2026
MLForSys2026 is an interdisciplinary workshop at NeurIPS 2026 that brings together researchers from machine learning and computer systems to advance the application of ML techniques—especially LLMs, multimodal models, and agentic workflows—to systems problems. It focuses on developing principled approaches beyond heuristic replacement, with emphasis on improving rigor, reproducibility, and trustworthiness in ML for Systems research.
Paper fit
Contribution paths
A strong submission should clearly identify its contribution and evaluate it appropriately.
Research areas in scope
01
ML for Systems Applications
Supervised, unsupervised, reinforcement learning, and generative AI (e.g., LLMs) applied to systems softwareRuntime systemsDistributed systemsSecurityCompilers, data structures, and code optimizationDatabasesComputer architecture, microarchitecture, and acceleratorsCircuit design and layoutInterconnects and NetworkingStorageData CentersProgramming LanguagesRepresentation learning for hardware and softwareOptimization of computer systems and softwareBenchmarking and evaluation methodologies for ML for SystemsDatasets and simulators for systems optimization and managementReliability, robustness, safety, and trustworthiness of ML-driven systemsSustainable and energy-efficient computingEmerging applications
02
LLMs and Agentic Workflows
Using LLMs and agentic workflows to address complex systems problems (e.g., program synthesis for hardware design, compiler autotuning, adaptive runtime optimization, system debugging, design-space exploration)Using ML to address emerging systems challenges in large-scale AI infrastructure (e.g., training and serving of LLMs, multimodal models, and agentic applications)Compiler partitioning strategies for distributed trainingMemory and compute allocationWorkflow orchestrationResource schedulingInference optimizationAutomated infrastructure managementEfficient utilization of heterogeneous accelerators
03
Methodologies and Infrastructure
Developing best practices, methodologies, benchmarks, datasets, simulators, and evaluation frameworks for ML for Systems research
Policies worth checking twice
- Submissions must be up to 4 pages (strict limit), excluding references and appendices.
- Additional material may be included in an optional appendix, but reviewers are not required to read it.
- Submissions must be in PDF format and follow the NeurIPS 2026 format.
- Submissions do not have to be anonymized.
- Accepted papers may be published elsewhere (no formal proceedings).
- Submission deadline: August 29, 2026, midnight anywhere on Earth.
Official sources
Compiled from the official call for papers. The organizers’ pages remain authoritative.
Last verified September 9, 2026