Steffen Herbold

A new study led by @sherbold.bsky.social from our university is investigating who is liable when #AI is misused to create child sexual abuse material. The authors reach a clear conclusion: "Anyone who develops AI must implement clear protective mechanisms – both technical and legal."

Announcing the #ICML2026 policy for self-ranking in reviews!
1. Authors rank their submissions
2. Reviews are submitted
3. The "Isotonic Mechanism" is run on rankings and review scores
4. Large discrepancies are flagged to ACs and SAC

CC @wjsu.bsky.social

Post: blog.icml.cc/2026/01/12/i...

This is not even considering the dilemma group leaders face when they are forced to rank their possibly very different, but still all high-quality publications of multiple group members ...

So authors wo submit a lot of low-quality papers will get special attention for their good papers if the scores are low.

Authors with a single paper will not get special attention when their scores are low.

Cool. I see no way how this could backfire 😐

Proud at my student Anamaria Mojica-Hanke for leading this work. It was a great collaboration with Thomas Goger (prosecutor at Bavaria's cybercrime unit), Brian Valerius (professor for AI in criminal law) and his student Svenja Wölfel.

tl;dr: The really bad content moderation of X/Grok is exactly what could lead to criminal prosecution for aiding and abetting CSAM generation. Slow take-down and making the content available via X adds more legal peril.

(Note: analysis based on German criminal law.)

Criminal Liability of Generative Artificial Intelligence Providers for User-Generated Child Sexual Abuse Material The development of more powerful Generative Artificial Intelligence (GenAI) has expanded its capabilities and the variety of outputs. This has introduced significant legal challenges, including gray a...

While the Grok-caused CSAM scandal is happening over on X, our work discussing the possible criminal liability of X (and others when publishing generative models) has been accepted the International Conference on AI Engineering.

The preprint is already online: arxiv.org/abs/2601.03788

Fun Christmas party of our research group yesterday. Fortunately, we found the exit!

We have 4 open PhD positions in the Future of Software Engineering (FUSE) lab! Topics:
- code review efficiency
- predictive software testing
- automated code refactoring
- engineering productivity metrics

(1/2)

YouTube video by Universität Passau Forschung mit Strahlkraft in die Region und darüber hinaus

Talente auf der Bühne, spannende Projekte, lebendiger internationaler Austausch und verborgene Talente: Rückblick auf die Forschungskommunikation 2025 u.a. mit @lingulist.de, @sherbold.bsky.social, @haeussler.bsky.social, @mgrani.bsky.social, @hedwigeisenbarth.bsky.social, @passaudpe.bsky.social:

The whos, whats, and whys of issues related to personal data and data protection in open-source projects on GitHub - Empirical Software Engineering Data protection regulations such as the General Data Protection Regulation (GDPR) in the European Union and the California Consumer Privacy Act (CCPA) in the US affect how software may handle the pers...

“The whos, whats, and whys of issues related to personal data and data protection in open-source projects on GitHub” by Anne Hennig, Lukas Schulte, Steffen Herbold, Oksana Kulyk, and Peter Mayer will be published in #EMSE! It examines discussions on personal data and data protection on #GitHub. 1/2

Passau study shows: AI passes as second corrector in exams Researchers at the University of Passau have had human examiners compete against OpenAI's ChatGPT – and were themselves surprised by some of the results. The study has been published in the renowned N...

🤖🎓 How good is #ChatGPT as a university correction assistant? Researchers from our university investigated this question – and were surprised by some of the results. The findings have been published in #ScientificReports @natureportfolio.nature.com

Original study: www.nature.com/articles/s41... 🧪

MAMUT: A Novel Framework for Modifying Mathematical Formulas for the Generation of Specialized Da...

Jonathan Drechsel, Anja Reusch, Steffen Herbold

Action editor: Hongsheng Li

https://openreview.net/forum?id=khODmRpQEx

#embeddings #notation #representations

Professor Tomas Sauer (standing) and Professor Christoph Heinzl demonstrate the visualization of a large industrial CT data set in a three-dimensional representation.

🚀 Shaping the future of 4D imaging

How to study the hidden dynamics of materials that transform themselves? Our university is part of the new #MSCA Doctoral Network #XCELERATE, pioneering 4D X-ray tomography methods.

Two fully funded doctoral positions are available at our university:

Sometimes reviewers still manage to surprise me.

A reviewer suggested, we should do a field study for something were we argue that is criminal.

We are now planning do address this and wondering if we the reviewer rather wants us to commit crimes or to become criminal investigators 🤨

And now some poor engineer at OpenAI has to design this emoji and add new training data. Good times.

Very interesting paper that shows that LLMs are good. But not good enough. I wish the following from the conclusion would have made it into the abstract:

😂

Just accepted at TMLR:

We found evidence of copyright violations by LLMs even when we ask questions that were not part of the training. Indeed, we found that the amount of memorized content was independent from the questions being part of the training or not.

openreview.net/forum?id=ddo...

Why language models hallucinate OpenAI’s new research explains why language models hallucinate. The findings show how improved evaluations can enhance AI reliability, honesty, and safety.

This just in: Leading AI firm discovers confidence thresholds. More on this exciting development in news at 11.

openai.com/index/why-la...

(Honestly, OpenAI!?)

Scientific impact and achievement, redefined:
Huge congrats to #Fraunhofer IIS on winning an #Emmy for their JPEG XS compression standard 🏆🎉 […]

re

(I miss IRC)

(Now I feel old)

Dear all,

please enjoy your complementary "European Professor goes on Holiday" message.

See you in September.

Yours sincerely,
A European Professor

Good news (for me!) my gender bias paper from 2023 still replicates with GPT-5.
Bad news (for everyone!) my gender bias paper from 2023 still replicates with GPT-5.
arxiv.org/pdf/2308.14921
hkotek.com/blog/gender-...

I wonder what my PhD students will think, once they discover that "someone" glued the three laws to the wall in the hallway. 🙃

Newton's Laws of Graduation, Part 2 - The Second Law

Newton's Laws of Graduation, Part 3 - The Third Law 😆

Success, a luxury problem, and its solution:
🎉 Our quiz is a huge success and incredibly popular on YouTube with now over 100,000 views.
😐 We cannot answer all the feedback and comments individually anymore.
😀 We write a follow up article to answer the most important questions.

Partial Colexifications Improve Concept Embeddings Arne Rubehn, Johann-Mattis List. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

It is official, our two long papers at #ACL2025 have now been published. Common work with Arne Rubehn (Concept Embeddings), and Frederic Blum and @sherbold.bsky.social (Automated Language Affiliation).

aclanthology.org/2025.acl-lon...
aclanthology.org/2025.acl-lon...

My debut as TV-Show moderator - now live on Youtube.

We had a lot of fun with how the five professors answered questions on topics ranging from 90's music, counting peas, size of Asian countries, etc.

The only drawback: it is only available in German.

P.S. The humans won.