3rd international Workshop on "AI + Astronomy: Models, Data, Discovery"

Oct 19, 2026, 8:30 AM → Oct 22, 2026, 8:30 PM Asia/Shanghai

Main Building (Zhejiang Lab)

Jian Wang (Zhejiang Lab), Jean-Paul Kneib (EPFL), Junhui Shi (Zhejiang lab)

We are delighted to invite you to the 3rd International Workshop on AI + Astronomy: Models, Data, Discovery. Over four days, we will explore how artificial intelligence is reshaping the field of astronomy. Building on the success of our first workshop in Hangzhou, China, and the second in Lausanne, Switzerland, we are proud to return to Hangzhou for this third edition.

The workshop will highlight the transformative role of large-scale AI models in enabling data-intensive scientific discovery.

The program will feature plenary and parallel sessions covering:

• Advances in AI and large models for astronomy
• Applications to spectral surveys, imaging, and time-domain data
• Computing, simulations, and data-intensive astronomy
• Instrumentation and frontier technologies
• International collaboration and open science

The main workshop will be held from 19 -22 Oct 2026 at the Zhejiang Lab in Hangzhou, China. Hosted by Zhejiang Lab in collaboration with EPFL Laboratory of Astrophysics. Leading scholars will gather to exchange ideas and explore the latest progress and future directions at the intersection of AI and astronomy.

Mon, October 19

8:30 AM Registration and Welcome Coffee

Opening Remarks

Convener: Jean-Paul Kneib (EPFL)

Keynote Speech

Convener: Junhui Shi (Zhejiang lab)

1 Title TBD

Speaker: Jian Wang (Zhejiang Lab)

2 Title TBD

Speaker: Jean-Paul Kneib (EPFL)

3 From Spectra to Scientific Action: Building Trustworthy AI for Astronomical Discovery

Large spectroscopic surveys provide an unprecedented foundation for astronomical discovery, yet scientific value does not emerge from data volume or model performance alone. Researchers must still reconcile heterogeneous data products, resolve scientific meaning, select appropriate analysis tools, assess uncertainty, and preserve the evidence required to reproduce a result.
This talk presents a service-oriented framework that turns multi-survey spectra into trustworthy scientific action. The framework begins with AI-ready spectral resources that preserve signal, uncertainty, quality flags, processing history, object identity, data versions, and provenance. An astronomical spectroscopy ontology then connects observations, physical quantities, celestial objects, analysis methods, and scientific evidence through explicit semantic constraints. Versioned scientific services provide validated capabilities for preprocessing, spectral measurement, parameter estimation, classification, redshift determination, and cross-survey retrieval.
Research agents operate on top of these resources and services. They interpret scientific questions, construct analysis plans, invoke appropriate tools, manage failures, compare competing hypotheses, and assemble evidence packages for expert review. Every result retains the data, ontology, model, and tool versions involved in its production.
Scientific cases are used as continuous system tests throughout requirements analysis, system design, development, integration, and acceptance. Examples include stellar parameter estimation, rare-object identification, galaxy redshift measurement, and multi-epoch analysis. Particular attention is given to rare objects and low-probability signals, which may be lost when knowledge systems rely recursively on model-generated content.
The goal is an open astronomical spectroscopy service in which data remain traceable, automated actions remain scientifically constrained, and every candidate discovery can be inspected, reproduced, challenged, and improved.

Speaker: Prof. ALi Luo (NAOC)

4 AI-Powered All-in-One Simulations for Precision Cosmology

Speaker: Cheng Zhao (Tsinghua University)

10:25 AM Group photo/Coffee break

Session 1：Astronomy Foundation model

Convener: Jean-Paul Kneib (EPFL)

5 Novel Approaches for Cosmological Survey Analysis

Speaker: Prof. Alexandre Refregier (ETH Zurich)

6 Towards AI Foundation Models for Large-Scale Structure Cosmology

Speaker: Prof. Tomasz Kacprzak (Fachhochschule Nordwestschweiz FHNW)

7 Machine Learning in Large-Scale Structure

Speaker: Yu Liu

8 Machine Learning for Cosmological Inference

Speaker: Leander Thiele (Kavli-IPMU)

9 Simulation-based inference of cosmology using neural compression of multi-probe maps

Speaker: Arne Thomsen (ETH Zurich)

12:10 PM Lunch

Session 1：Astronomy Foundation model

Convener: Tomasz Kacprzak (Fachhochschule Nordwestschweiz FHNW)

10 Fuzzy Dark Matter Simulations with Physics-Informed Generative Networks

Speaker: Prof. Emma Tolley (EPFL)

11 Scientific validation of deep generation model of galaxy images

Speaker: Shiyin Shen

12 Bridging JWST deep fields and wide-area surveys: A scale-free SOM for galaxy parameter estimation across heterogeneous surveys

Speaker: Qikai Liu (School of Astronomy and Space Science, Nanjing University)

13 Human-Machine Collaboration: Bridging the Last Mile Between Astronomers and Artificial Intelligence

Speaker: Nan Li (National Astronomical Observatories, Chinese Academy of Sciences)

14 How Much Compute Does Good Science Need? Co-Designing AI and Numerical Methods for Astronomy

Speaker: Denisa-Andreea Constantinescu (University of Basel)

2:45 PM Coffee Break

Session 2：Astronomical Data

Convener: Alexandre Refregie (ETH Zurich)

15 Two Frontiers of Radio Astronomy: AI for EHT and SKA Data

Speaker: Chin-Shin Chang (University of Geneva)

16 The Agentic Radiotelescope

Speaker: Adrien Devresse (Ecole Polytechnique Federale de Lausanne)

17 Semi-supervised representational learning for scalable radio source analysis

Speaker: Markus Bredberg (EPFL)

3:50 PM Coffee break

Session 2：Astronomical Data

18 AI for Astrophysical Visualization: New Developments in the VIRUP Project

Speaker: Florian Cabot (EPFL)

19 Code-specialized LLMs in HPC: a case study in radio-astronomy pipeline acceleration

Speaker: Hamza Chouh (EPFL)

20 AI based compression for Radio Astronomy

Speaker: Adrien Devresse (Ecole Polytechnique Federale de Lausanne)

4:55 PM Session discussion and wrap-up

6:00 PM Dinner

Tue, October 20

Session 3: PartA-AI and Spectroscopy of Galaxies

Convener: Emma Tolley (EPFL)

21 From JWST to ELT/MOSAIC: Deciphering Black Holes and Stellar Populations in the Early Universe

Speaker: Prof. Michaella Hirschmann (EPFL)

22 The Subaru Prime Focus Spectrograph (PFS) Cosmology Survey: overview and progress

Speaker: Jiaxi Yu (Kavli IPMU)

23 Diagnostics of Intermediate-Mass Black Holes and Ionising Sources

Speaker: Adele Plat (École Polytechnique Fédérale de Lausanne（EPFL）)

24 From Spectroscopic Redshift Errors to AI-Assisted Diagnostics

Speaker: Shengyu He (瑞士洛桑理工)

25 Rethinking Validation for Deep-learning Photometric Redshifts: The Impact of Redshift Distribution Shift

Speaker: 竹 陈 (上海师范大学)

26 On the hunt for XMPGs

Speaker: Timour Jestin (EPFL)

10:30 AM Coffee Break

Session 3: PartB-Astronomical Tools

Convener: Chao Liu (National Astronomical Observatories, CAS)

27 Illuminating the Dark Sector with Gravitational Lensing : from large space observatories programmes to whole sky surveys

Speaker: Mathilde Jauzac (IRAP)

28 LLM-Driven Automated Denoising Algorithm Search for KiDS DR5 Weak Lensing Shear Field

Speaker: Jiacheng Ding (Shanghai Astronomical Observatory, Chinese Academy of Sciences)

29 LensAgent: A Self Evolving Agent for Autonomous Physical Inference of Sub-galactic Structure

Speaker: Zihan Wang

30 Executable Scientific Verification with AI Agents: Formalizing Expert Reasoning in Astronomical Spectroscopy

Speaker: Haosong Wang (中国科学院上海天文台)

12:00 PM Lunch

Session 4

31 Emulating the complex galactic-scale orbital dynamics of LISA massive black hole pairs with normalizing flows

Speaker: Pedro R. Capelo (University of Zurich)

3:00 PM Coffee Break

Session 4

32 Accelerating Physics-Based Simulations with Deep Learning

Speaker: Dr Xuejian Jiang (Zhejiang Lab)

33 Discrete Tokenization of Continuous Astronomical Spectra for Foundation Models

Speaker: Dr Zhongyue Chen (Zhejiang Lab)

34 A Multimodal Foundation Model for Astronomical Object Search and Scientific QA

Speaker: Dr Kuang Mao (Zhejiang Lab)

6:00 PM Dinner

Wed, October 21

35 Hands-on Session 1

12:00 PM Lunch

36 Hands-on Session 2

3:00 PM Coffee Break

37 Hands-on Session 3

5:30 PM Dinner

Thu, October 22

38 Collaboration Discussion & Free talks

12:00 PM Lunch

1:30 PM Visits

1)Zhejiang Lab
2)Chengxi Sci-Tech Innovation Corridor Exhibition Hall
3)TBD

6:00 PM Dinner