Stefan Blücher

Join BIFOLD as a doctoral student in our Graduate School. Apply until February 03, 2025. Job ID: jobs.tu-berlin.de/en/job-posting… #jobvacancy #jobalert #ScienceJob #AI #research #Berlin

Our work on historical insights at scale using machine learning is now out in @ScienceAdvances! Very proud of this team effort, bridging disciplines and institutions—@MPIWG @TUBerlin @bifoldberlin @ml_tuberlin 📜science.org/doi/10.1126/sc…

#Job for #PhDCandidates: Develope and implement #XAI methods for ML models in Quantum Chemistry. BIFOLD is seeking a Research Associate in #ML for a project supervised by Dr. Stefan Chmiela and Dr. Shinichi Nakajima. Apply by 7th Nov '24 linkedin.com/jobs/view/4047… @ml_tuberlin

I have two vacant PhD positions in my group: 🏃 Machine Learning for Activity Trajectories s.gwdg.de/cSWx1D (Deadline Oct 20) 🧪 Machine learning for Medical Diagnostics s.gwdg.de/6NJfae (Deadline Oct 31) Looking forward to your applications!

✨ Our xMIL paper has been accepted at #NeurIPS2024 ✨ arXiv preprint: arxiv.org/abs/2406.04280

🚀Last-minute check-in to an academic career? We're seeking a PhD student to research the security and privacy of generative AI. Fully funded position in vibrant Berlin! Deadline is this Friday, September 20th. Apply now: #jobs" target="_blank" rel="nofollow noopener">mlsec.org/jobs.html#jobs @bifoldberlin @TUBerlin

@3blue1brown @jvielhaben @nstrodt This nicely algins with our MCD paper, in which we also focus on measuring angles between mult-dimensional spaces. openreview.net/forum?id=KxBQP…

Excerpt from a video I just posted on YouTube (about the multilayer perception layers in a transformer, and how LLMs may store facts).

🔊 New publication in @TmlrOrg Decoupling Pixel Flipping and Occlusion Strategy for Consistent #XAI Benchmarks, by @sbluecher @jvielhaben and @nstrodt together with @bifoldberlin and @FraunhoferHHI 🙏🏼 #MachineLearning #AI

Greatful to be published @TmlrOrg. Thanks to @nstrodt and @jvielhaben for their support. @bifoldberlin @ml_tuberlin @FraunhoferHHI

Decoupling Pixel Flipping and Occlusion Strategy for Consistent XAI Benchmarks Stefan Bluecher, Johanna Vielhaben, Nils Strodthoff. Action editor: Wei Liu. openreview.net/forum?id=bIiLX… #occlusion #features #feature

Our work on pixel flipping methodologies for #XAI methods got accepted at TMLR. Give our SRG measure a try the next time you benchmark your latest XAI method. Thanks @sbluecher @jvielhaben for the fun collaboration @XAI_Research @FraunhoferHHI @bifoldberlin

👩‍🔬Meet some of the BIFOLD researchers of the KI-Schnitzeljagd! Learn about AI while having fun. Don’t miss it! June 22, 2024 Fraunhofer HHI, Berlin langenachtderwissenschaften.de/programm/detai… #BehindTheScenes #AIResearch #LNDW2

Join BIFOLD and Machine Learning Group at @TUBerlin Berlin as a #PhD candidate. Conduct cutting-edge research in deep learning & explainable AI. Apply by June 14, 2024. jobs.tu-berlin.de/en/job-posting… @ml_tuberlin

Interested in higher-order #XAI Check out my recent package: github.com/bluecher31/hoo… It allows to calculater arbitrary interactions of features for #PredDiff and Shapley values. @trustworthy_ml @XAI_Research

Stefan Blücher is a member of the BIFOLD Graduate School. The next application deadline is closing soon: 26. Jan 2024. bifold.berlin/about-us/oppor… 10 open positions #PhD #ResearchJob #Berlin #ScienceJob x.com/bifoldberlin/s…

(1) grouping PF setups according to the R-OMS score clusters consistent MIF and LIF rankings. (2) decoupling reliability and insightfulness of PF benchmarks by combing MIF and LIF into the 'Symmetric Relevance Gap' measure.

Faithful attributions are measured by pixel flipping! But how trustworthy and consistent are the PF benchmarks? Read our new preprint to learn more! @nstrodt @jvielhaben @ml_tuberlin @bifoldberlin @FraunhoferHHI #XAI #diffusion arxiv.org/abs/2401.06654

@KirmizialtinL @GroupTuckerman Very intersting work. You might be interested in our recent work, which offers you the possibility to calculate computationally cheap higher-order attributions (similar to SHAP interactions). #br0370" target="_blank" rel="nofollow noopener">sciencedirect.com/science/articl…

Delighted to share our collaborative work with @GroupTuckerman, where we explored the remarkable potential of machine learning methods in identifying reaction coordinates. Check it out:pubs.aip.org/aip/jcp/articl…

🎙️ In todays #PhD seminar @sbluecher and Marco Morik presented their #research on „Assessing what really matters for occlusion-based #XAI“ and „#DiffusionModels for Inverse Problems“ Thanks guys for the interesting talks and discussions! #MachineLearning

Follow our lab to stay tuned for our most recent research highlights! #MachineLearning @bifoldberlin @TUBerlin