Riccardo De Santi

ETH AI Center PhD student. Exploration for Out-of-Distribution Discovery: from Theory to Molecules.

desanti_riccardo.jpg

OAT Y 14, ETH

Zurich, Switzerland

I am a Ph.D. student at ETH Zurich, advised by Andreas Krause, Niao He, and Kjell Jorner, and supported by the ETH AI Center and NCCR Catalysis. After visiting California Institute of Technology (Caltech), I continue working with Yisong Yue’s group and Frances H. Arnold’s lab to close the loop between generative discovery and chemical wet-lab validation. Recently, I was selected as a 2026 Rising Star in Data Science by UChicago, Stanford, Harvard, and UCSD. I serve as a research mentor for LeadTheFuture.

My research focuses on developing generative algorithms for discovery beyond the data — bridging flow and diffusion modeling, decision-making under uncertainty, and optimization, to enable new-to-nature discovery. Broadly, I aim to contribute to the foundations of a science of generative discovery: principled methods that move generative modeling beyond distribution matching and toward the discovery of new, valid, and useful structures, designs, and hypotheses.

This research program builds on my earlier work on the foundations of exploration in RL, which includes an Outstanding Paper Award at ICML with Marcello Restelli, and research visits with Michael Bronstein at the University of Oxford and Imperial College London on geometric and causal inductive biases for exploration.

Feel free to reach out if you wish to collaborate, exchange ideas, or seek project or thesis supervision.

Contacts:   rdesanti@ethz.ch   |   Google Scholar   |   Twitter   |   LinkedIn   |   Github

News

Sep 30, 2026 Selected as a 2026 Rising Star in Data Science by UChicago, Stanford, Harvard, UCSD.
Sep 28, 2026 Active Flow Expansion for Out-of-Distribution Discovery: from Theory to Molecules and On the Design Space of Discrete Diffusion Online Adaptation for Molecular Optimization (collab. with Genesis Molecular AI) accepted at NeurIPS 2026.
Sep 26, 2026 Our Next Generation of Discovery Loops session, co-organized with Samuel Kaski, Chris Holmes, Yee Whye Teh, and colleagues, has been accepted at the ELLIS UnConference 2026 in Paris.
Jun 30, 2026 Gave a series of talks on Foundations of Generative Discovery Beyond the Data via Flow and Diffusion Models at Caltech, UC Berkeley, UCLA, UCSD, UC Irvine, Stanford.
Apr 30, 2026 Three works on Flows for Discovery accepted at ICML 2026: (1) A Unified Density Operator View of Flow Control and Merging, (2) Constrained Flow Optimization via Sequential Fine-Tuning for Molecular Design, and (3) Efficient Tail-Aware Generative Optimization via Flow Model Fine-Tuning.
Apr 15, 2026 A Unified Density Operator View of Flow Control and Merging accepted with Oral presentation at the Workshop on Real-World Constrained and Preference-Aligned Flow and Diffusion-Based Models at ICLR 2026.
Jan 26, 2026 Three works on Flows for Discovery accepted at ICLR 2026: (1) Verifier-Constrained Flow Expansion for Discovery Beyond the Data, (2) Riemannian Optimization via Pre-trained Diffusion Models, and (3) Value Matching: Scalable and Gradient-Free Reward-Guided Flow Adaptation.
Oct 15, 2025 Co-designed with Andreas Krause the new Diffusion Generative Models module for ETH Zurich’s Probabilistic AI course (~500 students).
Oct 6, 2025 Constrained Flow Optimization via Sequential Fine-Tuning for Molecular Design awarded Oral at the Frontiers in Probabilistic Inference Workshop at NeurIPS 2025.
Sep 18, 2025 Flow Density Control: Generative Optimization Beyond Entropy-Regularized Fine-Tuning has been accepted as Spotlight at NeurIPS 2025
Jul 8, 2025 Flow Density Control: Generative Optimization Beyond Entropy-Regularized Fine-Tuning accepted as an Oral at the Workshop on Generative AI and Biology at ICML 2025

Selected Publications

  1. NeurIPS
    Active Flow Expansion for Out-of-Distribution Discovery: from Theory to Molecules
    Riccardo De Santi, Bruce Lee, Cristian Perez Jensen, and 6 more authors
    Advances in Neural Information Processing Systems (NeurIPS), 2026
    Bibtex PDF Code Website
  2. ICMLOral Presentation
    A Unified Density Operator View of Flow Control and Merging
    Riccardo De Santi, Malte Franke, Ya-Ping Hsieh, and 1 more author
    International Conference on Machine Learning (ICML), 2026
    Oral at Real-World Constrained and Preference-Aligned Flow and Diffusion-Based Models Workshop at ICLR 2026
    Bibtex PDF
  3. ICMLOral Presentation
    Constrained Flow Optimization via Sequential Fine Tuning for Molecular Design
    Sven Gutjahr*, Riccardo De Santi*, Luca Schaufelberger*, and 2 more authors
    International Conference on Machine Learning (ICML), 2026
    Oral at Frontiers in Probabilistic Inference Workshop at NeurIPS 2025
    Bibtex PDF Code
  4. ICLR
    Verifier-Constrained Flow Expansion for Discovery Beyond the Data
    Riccardo De Santi*, Kimon Protopapas*, Ya-Ping Hsieh, and 1 more author
    International Conference on Learning Representations (ICLR), 2026
    Bibtex PDF Code
  5. NeurIPSSpotlight and Oral Presentation
    Flow Density Control: Generative Optimization Beyond Entropy-Regularized Fine-Tuning
    Riccardo De Santi, Marin Vlastelica, Ya-Ping Hsieh, and 3 more authors
    Advances in Neural Information Processing Systems (NeurIPS), 2025
    Spotlight at NeurIPS 2025 and Oral at Workshop on Generative AI and Biology at ICML 2025
    Bibtex PDF Code Website
  6. ICML
    Provable Maximum Entropy Manifold Exploration via Diffusion Models
    Riccardo De Santi*, Marin Vlastelica*, Ya-Ping Hsieh, and 3 more authors
    International Conference on Machine Learning (ICML), 2025
    Bibtex PDF Code
  7. ICMLOutstanding Paper
    The Importance of Non-Markovianity in Maximum State Entropy Exploration
    Mirco Mutti*, Riccardo De Santi*, and Marcello Restelli
    International Conference on Machine Learning (ICML), 2022
    Outstanding Paper Award at ICML 2022
    Bibtex Talk PDF

Selected Talks

2026
Active Generative Models for Out-of-Distribution Discovery: from Theory to Molecules
UNIST, South Korea — post-ICML Workshop hosted by the Active Generative Agents Lab
Slides
2026
Foundations of Generative Discovery Beyond the Data via Flow and Diffusion Models
Caltech, UC Berkeley, UC Irvine, UCLA, UCSD, Stanford
Slides
2026
Generative Discovery Beyond the Data with Flow Models: from Algorithms to Molecules
Genesis Molecular AI, San Mateo, California
2024
Towards Generative Decision-Making for Scientific Discovery
ETH AI Center, Zurich, Switzerland
2023
Towards Convex Reinforcement Learning for the Real-World
ETH AI Center, Zurich, Switzerland
2022
The Role of Causal Structure in Unsupervised Reinforcement Learning
Technische Universität Dresden, Dresden, Germany — only M.Sc. student as RL workshop speaker