Shunchang Liu
Portrait

Hi! I'm Shunchang (顺昌), a PhD student in the Theory of Machine Learning (TML) Lab at EPFL, advised by Prof. Nicolas Flammarion and co-supervised by Prof. Francesco Croce. Before starting my PhD, I completed a joint Master's degree in Cybersecurity at EPFL and ETH Zurich.

My research focuses on AI Alignment. I'm currently most interested in the underlying principles of alignment, that is, understanding why models exhibit misaligned behaviors that deviate from human expectations, and how to control them from first principles. I've also worked on AI safety and copyright.

I'm honored to collaborate closely with Prof. Boi Faltings at EPFL, and Prof. Andreas Krause and Prof. Florian Tramer at ETH Zurich.

Google Scholar GitHub LinkedIn CV
Education
  • École Polytechnique Fédérale de Lausanne
    Ph.D. in Computer Science
    Sep. 2026 – present
  • École Polytechnique Fédérale de Lausanne
    M.S. in Cybersecurity
    Sep. 2024 – Aug. 2026
  • Eidgenössische Technische Hochschule Zürich
    M.S. in Cybersecurity
    Sep. 2024 – Aug. 2026
Internship
News
2026
Honored to receive the EPFL EDIC PhD Fellowship!
Jun 10
Excited to start my Master Thesis at ETH SPY Lab! [Read more]
Mar 01
2025
One paper was accepted by NeurIPS @GenAI4Health! [Read more]
Sep 30
Two reports I contributed to, Frontier AI Risk Management Framework [Read more] and Responsible Innovation in AI × Life Sciences [Read more], were launched at WAIC 2025!
Jul 28
One paper was accepted by ACM MM! [Read more]
Jul 04
Selected Publications
Preference Instability in Reward Models: Detection and Mitigation via Sparse Autoencoders
Preference Instability in Reward Models: Detection and Mitigation via Sparse Autoencoders

Shunchang Liu, Xin Chen, Belen Martin Urcelay, Francesco Croce

ICML @CompLearn Workshop & @Mech Interp Workshop, 2026

[Paper] [Code]

Preference Instability in Reward Models: Detection and Mitigation via Sparse Autoencoders

Shunchang Liu, Xin Chen, Belen Martin Urcelay, Francesco Croce

ICML @CompLearn Workshop & @Mech Interp Workshop, 2026

[Paper] [Code]

CopyJudge: Automated Copyright Infringement Identification and Mitigation in Text-to-Image Diffusion Models
CopyJudge: Automated Copyright Infringement Identification and Mitigation in Text-to-Image Diffusion Models

Shunchang Liu*, Zhuan Shi*, Lingjuan Lyu, Yaochu Jin, Boi Faltings (* equal contribution)

ACM International Conference on Multimedia (MM), 2025

[Paper] [Code] [Video]

Thanks for Anderson's insightful commentary and sharing (1, 2) by the Korea Copyright Commission.

CopyJudge: Automated Copyright Infringement Identification and Mitigation in Text-to-Image Diffusion Models

Shunchang Liu*, Zhuan Shi*, Lingjuan Lyu, Yaochu Jin, Boi Faltings (* equal contribution)

ACM International Conference on Multimedia (MM), 2025

[Paper] [Code] [Video]

Thanks for Anderson's insightful commentary and sharing (1, 2) by the Korea Copyright Commission.

Harnessing Perceptual Adversarial Patches for Crowd Counting
Harnessing Perceptual Adversarial Patches for Crowd Counting

Shunchang Liu*, Jiakai Wang*, Aishan Liu, Yingwei Li, Yijie Gao, Xianglong Liu, Dacheng Tao (* equal contribution)

ACM SIGSAC Conference on Computer and Communications Security (CCS), 2022

[Code] [Paper] [Video]

Harnessing Perceptual Adversarial Patches for Crowd Counting

Shunchang Liu*, Jiakai Wang*, Aishan Liu, Yingwei Li, Yijie Gao, Xianglong Liu, Dacheng Tao (* equal contribution)

ACM SIGSAC Conference on Computer and Communications Security (CCS), 2022

[Code] [Paper] [Video]

All publications