Lightnews — Scholar-powered news

BIFOLD Berlin Institute for the Foundations of Learning and Data @bifold.berlin · 2d

Tagesspiegel named Dr. Stefan Chmiela among Berlin’s 100 most important minds in science 2025.

www.bifold.berlin/news-events/...

@tuberlin.bsky.social @tagesspiegel.de #MLsky

Chmiela named among Berlin’s Top 100 Scientists 2025

BIFOLD research group lead Dr. Stefan Chmiela has been named one of the “100 Most Important Minds in Berlin Science 2025” by Tagesspiegel. His research uses explainable AI to model molecular processes...

www.bifold.berlin

1 2

BIFOLD Berlin Institute for the Foundations of Learning and Data @bifold.berlin · 6d

👾 PhD Position in Database Systems & Big Data Analytics
Focus: database systems, distributed and parallel computing, and big data analytics.

🗓️ Application deadline: November 15, 2025
📍 TU Berlin, BIFOLD – Data Management
www.jobs.tu-berlin.de/en/job-posti...

BIFOLD Berlin Institute for the Foundations of Learning and Data @bifold.berlin · 6d

🧠 PhD Position in Graph Neural Networks & Explainable AI
Focus:Graph Neural Networks (GNNs), transformer models, and explainable AI (XAI).

🗓️ Application deadline: October 17, 2025
📍 TU Berlin, BIFOLD – Machine Learning
www.jobs.tu-berlin.de/en/job-posti...

1

BIFOLD Berlin Institute for the Foundations of Learning and Data @bifold.berlin · 6d

Two PhD Opportunities at BIFOLD / #TUBerlin :

1x #ML #XAI #GNN; Group suvised by KR Müller
1x #DBS #BDA; Group supervised by V Markl

www.bifold.berlin/about-us/opp...

#PhDSky #fundedPhD #doctoral #researcher #sciencejobs #Stemjobs #academicjobs #MLSky @datasciencejobscom.bsky.social

1 1 1

Reposted by BIFOLD Berlin Institute for the Foundations of Learning and Data

Nils Feldhus @nfel.bsky.social · 6d

🔍 Are you curious about uncovering the underlying mechanisms and identifying the roles of model components (neurons, …) and abstractions (SAEs, …)?

We provide the first survey of concept description generation and evaluation methods.

Joint effort w/ @lkopf.bsky.social

📄 arxiv.org/abs/2510.01048

Overview of descriptions for model components (neurons, attention heads) and model abstractions (SAE features, circuits).

1 3 17

Reposted by BIFOLD Berlin Institute for the Foundations of Learning and Data

Max Planck School of Cognition @mps-cognition.bsky.social · Sep 4

The new application cycle for our fully funded international graduate program has just started. You can now apply via our website, sign up for a Q&A, or participate in the Applicant Support Program cognition.maxplanckschools.org/en ! 👍🏻🧠👏🏾#passionforscience, #maxplanckschools

1 15 14

BIFOLD Berlin Institute for the Foundations of Learning and Data @bifold.berlin · 13d

github.com/David-Drexli...

GitHub - David-Drexlin/MeDi

Contribute to David-Drexlin/MeDi development by creating an account on GitHub.

github.com

BIFOLD Berlin Institute for the Foundations of Learning and Data @bifold.berlin · 13d

📄 Paper: MeDi: Metadata-Guided Diffusion Models for Mitigating Biases in Tumor Classification
👩‍🔬 Authors: David Jacob Drexlin, Jonas Dippel, Julius Hense, Niklas Prenißl, Grégoire Montavon, Frederick Klauschen, Klaus-Robert Müller
🔗 Read the paper papers.miccai.org/miccai-2025/...

MeDi: Metadata-Guided Diffusion Models for Mitigating Biases in Tumor Classification

Abstract Deep learning models have made significant advances in histological prediction tasks in recent years. However, for adaptation in clinical practice, their lack of robustness to varying conditi...

papers.miccai.org

1

BIFOLD Berlin Institute for the Foundations of Learning and Data @bifold.berlin · 13d

BIFOLD researcher Jonas Dippel and David Jacob Drexlin will present a paper during the Friday morning poster session at MICCAI2025.

📄 Paper: MeDi: Metadata-Guided Diffusion Models for Mitigating Biases in Tumor Classification

#ML #MLSky @tuberlin.bsky.social

1 4

Reposted by BIFOLD Berlin Institute for the Foundations of Learning and Data

Andreas Wagner @anwagnerdreas.hcommons.social.ap.brid.gy · 14d

At the @bifold.berlin conference "AI-based methods in the humanities", I have just attended a great talk by Seid Muhie Yimam of Hamburg University who confirmed my impression that there is a kind of momentum in this area at the moment. He mentioned many datasets, publications and shared tasks on […]

Original post on hcommons.social

hcommons.social

1 3 1

BIFOLD Berlin Institute for the Foundations of Learning and Data @bifold.berlin · 14d

On Sep 25, 2025 Prof. Begüm Demir will deliver a keynote lecture at the GAIA 2025 Symposium: "From Efficient Foundation Models to AI Agents for Foundation Model Recommendations to Advance Earth Observation"

gaia.insait.ai

#earthobservation(s) #earthobs #remotesensing #AI4EO #GISchat

GAIA Symposium 2025

Geospatial AI and Applications with Foundation Models (GAIA) Symposium 2025 | 24-25 September 2024, Sofia, Bulgaria

gaia.insait.ai

1 4

BIFOLD Berlin Institute for the Foundations of Learning and Data @bifold.berlin · 15d

We’ve just kicked off our workshop “AI-based Methods for the Humanities”. Judging by the happy faces in the group photo below 👇, this is going to be a fascinating exchange of ideas!

www.bifold.berlin/news-events/...

#DH @mpiwg.bsky.social @tuberlin.bsky.social @ecdigitalfuture.bsky.social

2 6

Reposted by BIFOLD Berlin Institute for the Foundations of Learning and Data

Fischerdata @fischerdata.bsky.social · 15d

Attending a conference organised by @bifold.berlin on „AI based methods for the humanities“ I learn how computer scientists explore the way AI operates in order to improve their results. Like Nils Feldhus on the topic of explanation.

1 1 5

BIFOLD Berlin Institute for the Foundations of Learning and Data @bifold.berlin · 16d

🎤 About the speaker: Prof. Dr. Ingo Scholtes ( @uni-wuerzburg.de ) leads research at the intersection of network science, deep graph learning & computational social science. His work has appeared in Nature Physics and beyond.

1

BIFOLD Berlin Institute for the Foundations of Learning and Data @bifold.berlin · 16d

Applications range from social and technical systems to biology. 🌐 By modeling temporal-topological patterns, De Bruijn GNNs improve how we learn from time series data in complex networks.

1

BIFOLD Berlin Institute for the Foundations of Learning and Data @bifold.berlin · 16d

Enter higher-order De Bruijn Graphs: a tool to understand how temporal order impacts causal structures in networks. theoretical foundation for De Bruijn Graph Neural Networks — a new, time-aware deep learning architecture. #MachineLearning

1 1

BIFOLD Berlin Institute for the Foundations of Learning and Data @bifold.berlin · 16d

Not just who is connected matters — but when and in which order. ⏳ The arrow of time shapes how nodes influence each other in dynamic networks. This has big implications for graph analytics & deep learning. #AI #GraphLearning

1

BIFOLD Berlin Institute for the Foundations of Learning and Data @bifold.berlin · 16d

Only 3 days to go! 🚀 Join the BIFOLD Colloquium in Berlin on Sept 25, 4pm. Ingo Scholtes will speak on: Machine Learning for Complex Networks: From GNNs to Causal Graph Learning.
📍TU Berlin @tuberlin.bsky.social

www.bifold.berlin/news-events/...

#DH #DigitalHumanities #Digi-Hum

BIFOLD Colloquium 10/2025

In the context of the BIFOLD workshop "AI-based Methods for the Humanities", which brings together AI and the humanities to model historical, cultural, and linguistic phenomena, Prof. Dr. Ingo Scholte...

www.bifold.berlin

1 1

Reposted by BIFOLD Berlin Institute for the Foundations of Learning and Data

Laura Kopf @lkopf.bsky.social · 19d

Happy to share that our PRISM paper has been accepted at #NeurIPS2025 🎉

In this work, we introduce a multi-concept feature description framework that can identify and score polysemantic features.

📄 Paper: arxiv.org/abs/2506.15538

#NeurIPS #MechInterp #XAI

1 3 25

Reposted by BIFOLD Berlin Institute for the Foundations of Learning and Data

Konrad Rieck 🌈 @rieck.mlsec.org · 23d

Got some hot research cooking? 🔥

The @satml.org paper deadline is just 9 days away. We are looking forward to your work on security, privacy, and fairness in machine learning.

👉 satml.org/call-for-pap...
⏰ Sep 24

3 4

BIFOLD Berlin Institute for the Foundations of Learning and Data @bifold.berlin · 22d

Study: Seeing through: Analyzing and Attacking Virtual Backgrounds in Video Calls

Felix Weißberg, Jan-Malte Hilgefort, Steve Grogorick, Daniel Arp, Thorsten Eisenhofer, Martin Eisemann, Konrad Rieck.

Proceedings of the 34th USENIX Security Symposium, 2025. @tuberlin.bsky.social @tuwien.at

1 2

BIFOLD Berlin Institute for the Foundations of Learning and Data @bifold.berlin · 22d

Many of us assume that clicking “blur” or activating a virtual background protects our privacy. This research shows: that belief is misleading. The safest option is still the simplest one: prepare your real environment for calls instead of relying solely on digital masking.

1 1 2

BIFOLD Berlin Institute for the Foundations of Learning and Data @bifold.berlin · 22d

The team developed an attack that exposed 53% more background pixels than previous methods, showing just how much information can unintentionally slip through.

1 1 1

BIFOLD Berlin Institute for the Foundations of Learning and Data @bifold.berlin · 22d

These fragments are often invisible to the human eye, but when collected over time, they can be used to reconstruct rooms and even identify personal objects.

1 1 1

BIFOLD Berlin Institute for the Foundations of Learning and Data @bifold.berlin · 22d

Virtual backgrounds leak tiny image fragments of the real environment – so-called pixel leaks.

These fragments are often invisible to the human eye, but when collected over time, they can be used to reconstruct rooms and even identify personal objects.

1 1 1