Felix Friedrich

About Me

My research interests lie in inference and post-training techniques for generative models. I focus on multimodal models, in particular diffusion and flow-based architectures, and work on improving their capabilities and controllability.

Beyond generation research, I am deeply interested in the safety of multimodal models. As these systems become more powerful and widely deployed, ensuring they behave responsibly and do not produce harmful or biased outputs is critical. My work addresses both the advancement and the responsible deployment of generative AI.

Research Topics: Multimodal AI · Diffusion/Flow Models · Inference & Post-Training · AI Safety · AI Alignment

Timeline

2025 – present

Postdoctoral Researcher at Meta FAIR, Montreal, Canada. Working with Michal Drozdzal, Adriana Romero Soriano, and Luke Zettlemoyer.

2024 – 2025

× Researcher and Co-lead at Lab1141, AlephAlpha × TU Darmstadt, Germany.

2021 – 2025

× PhD student with Prof. Kristian Kersting at Machine Learning Lab, TU Darmstadt, and 3AI, hessian.AI, Germany.

2020

Erasmus+ at Chalmers University of Technology, Gothenburg, Sweden.

2019 – 2021

M.Sc. in Computer Science (minor in Psychology), TU Darmstadt, Germany.

2018 – 2021

M.Sc. (with honors) in Autonomous Systems, TU Darmstadt, Germany.

2017

Research internship on intelligent autonomous driving systems at IAV, Volkswagen Group, Germany.

2014 – 2017

B.Sc. in Electrical Engineering, TU Dortmund, Germany.

Selected Publications

For a full list, see my Google Scholar profile.

Inference-time Physics Alignment of Video Generative Models with Latent World Models

J Yuan, X Zhang*, F Friedrich*, N Beltran-Velez*, M Hall, R Askari-Hemmat, ...

CVPR 2026

arXiv Code

Measuring and Guiding Monosemanticity

R Härle*, F Friedrich*, M Brack, S Wäldchen, B Deiseroth, P Schramowski, ...

NeurIPS 2025 (Spotlight)

arXiv NeurIPS Code

Multilingual Text-to-Image Generation Magnifies Gender Stereotypes and Prompt Engineering May Not Help You

F Friedrich, K Hämmerl, P Schramowski, M Brack, J Libovicky, K Kersting, ...

ACL 2025

arXiv ACL Code

Evaluating the Social Impact of Generative AI Systems in Systems and Society

I Solaiman, Z Talat, W Agnew, L Ahmad, D Baker, SL Blodgett, C Chen, ...

Oxford Handbook on the Foundations and Regulation of Generative AI 2025

arXiv Oxford

LlavaGuard: An Open VLM-based Framework for Safeguarding Vision Datasets and Models

L Helff*, F Friedrich*, M Brack*, K Kersting, P Schramowski

ICML 2025

arXiv PMLR Website

Aurora-M: Open Source Continual Pre-training for Multilingual Language and Code

T Nakamura, M Mishra, S Tedeschi, Y Chai, JT Stillerman, F Friedrich, ...

COLING 2025

arXiv ACL Model (HF)

LEdits++: Limitless Image Editing using Text-to-Image Models

M Brack*, F Friedrich*, K Kornmeier*, L Tsaban, P Schramowski, ...

CVPR 2024

arXiv CVPR Code Website

ALERT: A Comprehensive Benchmark for Assessing Large Language Models' Safety through Red Teaming

S Tedeschi, F Friedrich, P Schramowski, K Kersting, R Navigli, H Nguyen, ...

Online Workshop on Red Teaming Generative AI Models 2024

arXiv Code & Dataset

Learning by Self-Explaining

W Stammer, F Friedrich, D Steinmann, M Brack, H Shindo, K Kersting

TMLR 2024

arXiv TMLR

SEGA: Instructing Text-to-Image Models using Semantic Guidance

M Brack, F Friedrich, D Hintersdorf, L Struppek, P Schramowski, ...

NeurIPS 2023

arXiv NeurIPS Demo

MultiFusion: Fusing Pre-trained Models for Multi-lingual, Multi-modal Image Generation

M Bellagente, M Brack, H Teufel, F Friedrich, B Deiseroth, C Eichenberg, ...

NeurIPS 2023

arXiv NeurIPS Code

Exploiting Cultural Biases via Homoglyphs in Text-to-Image Synthesis

L Struppek, D Hintersdorf, F Friedrich, P Schramowski, K Kersting

JAIR 2023

arXiv JAIR Code

A Typology for Exploring the Mitigation of Shortcut Behaviour

F Friedrich, W Stammer, P Schramowski, K Kersting

Nature Machine Intelligence 2023

arXiv Nature MI Code

Teaching

Supervised courses at TU Darmstadt with Prof. Dr. Kristian Kersting: