Short Paper
Up to 4 pages with unlimited references and appendices; double-blind review; non-archival; accepted papers presented as posters or oral/lightning talks.
Machine Learning/CORE Unranked
FMSD @ ICML 2026
2nd ICML Workshop on Foundation Models for Structured Data
The FMSD @ ICML 2026 workshop focuses on foundation models for structured data, specifically tabular and time-series data, aiming to unify research communities around shared challenges in data curation, scaling, evaluation, and real-world deployment. It emphasizes multimodal approaches, synthetic data generation, benchmarks that mitigate contamination, and comparisons between purpose-built models and LLM/agent-based methods on predictive structured tasks.
Key deadlines
Verified September 9, 2026Full paper
May 9, 2026
AoE
Workshop timeline
Submission and decisions
Full paperKey deadline
May 9, 2026 · AoE
Paper fit
Contribution paths
A strong submission should clearly identify its contribution and evaluate it appropriately.
Research areas in scope
01
Building Foundation Models for (Multimodal) Structured Data
Zero-shot inference and in-context learning on unseen dataNovel architectures and scaling insightsMultimodal structured foundation models integrating tabular/time-series with text and imagesCross-modality transfer (e.g., tabular to time-series)
02
Datasets and Synthetic Data Generation Methods
High-quality large-scale structured datasetsSynthetic data generation schemes for pretrainingData provenance and privacy auditing
03
Benchmarks
Unified evaluation protocols for structured foundation modelsInference throughput and memory usageScalability and data memorizationContamination and overfitting mitigation
04
Alternative Paradigms (LLMs and Agents)
LLM/agent-based evaluation on predictive tabular/time-series tasksCost/latency and numerical reliability reportingScaling trade-offs: sample/compute efficiency vs. purpose-built modelsInterpretability, efficiency, and reliability in structured settings
05
Applications of Foundation Models for Structured Data
Real-world applications in healthcare, finance, climate modeling, and supply chainDomain adaptation and model reliabilityData privacy and ethical considerationsFairness and bias mitigationDomain-specialized foundation models
Policies worth checking twice
- Submissions must be double-blind and properly anonymized.
- Papers exceeding 4 pages will be desk-rejected.
- Dual submission to ICML 2026 main conference is not allowed.
- Papers previously published in journals or conferences must be reasonably extended.
- No rebuttal phase; decisions based solely on submission and reviews.
- Author list cannot be changed after submission.
- Workshop is non-archival; accepted papers are available on OpenReview but not formally published.
- Camera-ready deadline is a best-effort suggestion with no enforcement.
Official sources
Compiled from the official call for papers. The organizers’ pages remain authoritative.
Last verified September 9, 2026