About - Registration - Schedule - Keynote Speakers - Keynote Abstracts - Accepted Contributions - Important Dates - Call for Papers
About
This will be the 5th-year edition of the ML4Molecules workshop organized by the ELLIS research program on “Machine Learning for Molecule Discovery”. This research program brings together a pan-European community dedicated to accelerating molecular discovery through cutting-edge AI. The meeting will serve as a key opportunity for members and collaborators to present their latest research in the area, exchange insights, coordinate research efforts, and align on strategic goals. ML4Molecules 2025 will focus on the latest developments in generative models, large language models (LLMs), and scalable AI systems for applications in chemistry, drug discovery, and materials science. We will explore topics such as: 1) Foundation and diffusion models for molecule generation, 2) Cross-modal and cross-domain learning (e.g., text, structure, experimental data), 3) AI-driven synthesis planning and reaction prediction, 4) Interpretable and data-efficient ML approaches. 5) Practical considerations in deploying ML in real-world molecular pipelines, etc. In addition to technical sessions, this edition will host the annual meeting of the ELLIS Program on Machine Learning for Molecule Discovery. Building on the progress of previous editions — from foundational methods (2021), critical assessments (2022–2023), to the LLM and foundation model revolution (2024) — ML4Molecules 2025 will foster rigorous, interdisciplinary dialogue and chart new directions for impactful molecular AI research.
Registration
TBA
Schedule
Keynote Speakers
Important dates
- Submission Date for Workshop Contributions: October 15, 2025, 23:59 UTC (extended)
- Accept/Reject Notification Date: October 31, 2025
- December 2, 2025: Workshop at the ELLIS Unconference at Copenhagen
Call for papers
We invite submissions on machine learning for molecules and materials with a special focus on generative models and large language models (LLMs) — from representation and property prediction to end‑to‑end discovery workflows and lab integration. This workshop is part of the ELLIS UnConference on December 2, 2025 (Copenhagen), co‑located with NeurIPS week.
Scope & Topics (include but not limited to)
- Generative modeling for molecules/materials (diffusion, normalizing flows, autoregressive, energy‑based) with goal‑directed or multi‑objective optimization
- LLMs and agents for chemistry & materials: code‑generation for simulation/analysis, autonomous experiment planning, protocol extraction, and tool use (e.g., retrosynthesis, docking, DFT/MD pipelines)
- Multimodal and structured reasoning: text‑to‑molecule/material, text+graph+structure fusion, retrieval‑augmented generation with chemical databases and knowledge graphs
- Property prediction & simulation surrogates: equivariant GNNs, learned interatomic potentials, speeding MD/MC/DFT, uncertainty quantification & robustness
- Data‑efficient learning: active learning, few‑/self‑/weakly supervised methods, synthetic data and simulation‑augmented training
- Reaction modeling & synthesis planning, catalyst and materials discovery, inverse design across drug, agrochemical, and energy applications
- Evaluation & safety: benchmarks, reproducibility, alignment of scientific LLMs, bias/risk assessment, and responsible deployment in discovery settings
Submission Instructions
- Platform: Openreview
- Format: PDF using the NeurIPS style (main text max 5 pages, excluding references and appendices).
- OR: Extended abstracts with up to 2 pages (PDF free form).
- Anonymity: Dual‑anonymous review—please anonymize your manuscript.
Presentation & Publication Policy
- Accepted papers will be presented as posters; a subset may be selected for spotlight talks.
- This workshop is non‑archival. Workshop papers often reflect ongoing work and will not be treated as final versions of record.
Accepted contributions (poster)
Best Paper Award
Sponsors
Organizing Committee and Contact
Chairs: Nadine Schneider, Francesca Grisoni, and Jose Miguel Hernandez Lobato
Organizing committee:
Contact: ml4molecules@ml.jku.at