🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

How can we use AI to automate our #rail transportation, when we all know how bad mobile data service can be on the train? In #BerDiBa, our scientists are using efficient computing to work on the future of #mobility. Learn more here: hhi.fraunhofer.de/en/departments…

Check out our paper "Reactive Model Correction: Mitigating Harm to Task-Relevant Features via Conditional Bias Suppression" at the #CVPR2024 #SAIAD workshop on June 18! We introduce a targeted, training-free model correction method that minimizes collateral damage within models.

New paper alert! With EMPRT and SMPRT we fix some glaring issues with the original Model Parameter Randomization Test (MPRT). Visit our talk at the 2nd XAI World Conference in Valetta, Malta, on July 17th! more info & paper: tinyurl.com/lnkdn-smprt-em…

Our latest paper is out! On Tuesday, 17 July, we will present our latest work "A Fresh Look at Sanity Checks for Saliency Maps" at the 2nd XAI World Conference in Valetta, Malta. I hope to see you there! Read the paper: link.springer.com/chapter/10.100… Code: github.com/annahedstroem/…

🚨 AttnLRP is at #ICML! Explain your transformers in a fast, efficient and best possible way! 🚨 more info: tinyurl.com/lnkdn-attnlrp paper: proceedings.mlr.press/v235/achtibat2… code: github.com/rachtibat/LXT

Go meet our #AttnLRP - Team at #ICML -- Reduan Achtibat, (Sayed) Erfan Hatefi and Aakriti Jain, right now! Discuss how to best (ie, fast, cheap, faithful!) explain your transformer-based model during Poster Session 2!

I'm very excited to share some great news: the paper I supervised has been accepted to #NeurIPS2024. Huge congratulations to my student and soon-to-be PhD candidate Laura Kopf on her first lead-author publication—what an incredible way to start a scientific journey!

🚨 New paper alert! 🚨 Excited to share Quanda—a toolkit for evaluating data attribution methods, accepted at #NeurIPS ATTRIB workshop!🎉 A collaboration between Fraunhofer HHI & BIFOLD. 🔗 Paper: arxiv.org/abs/2410.07158 🔗 Code: github.com/dilyabareeva/q…

We (Dilyara Bareeva, Galip Ümit Yolcu, Niklas Schmolenski, Thomas Wiegand, Wojciech Samek, Sebastian Lapuschkin + me) just launched QUANDA — a training data attribution TDA software Built for researchers curious to apply/ develop/ evaluate TDA methods GitHub repo: github.com/dilyabareeva/q…

✨Introducing quanda: an open-source library for benchmarking training data attribution (TDA) methods in PyTorch! Quanda offers a user-friendly interface for ⚖️ evaluating attributions and 📊 benchmarking TDA methods across diverse metrics.

Congrats to Fraunhofer HHI #XAI experts Wojciech Samek & Sebastian Lapuschkin! @Handelsblatt has listed them among the most cited #AI researchers in Germany, and Fraunhofer HHI as one of the most important #AIhubs in the country. 👉hhi.fraunhofer.de/kiforschungfra…

🚨 New paper alert! 🚨 We’re excited to share our latest work on interpretability evaluation: "Evaluating Interpretable Methods via Geometric Alignment of Functional Distortions" 📜 Accepted at TMLR 🎉 🔥 Survey certification 🔥 📖 Read: openreview.net/pdf?id=ukLxqA8…

Huge thanks again to our co-authors! Anna Hedström Philine Bommer Tom Burns Sebastian Lapuschkin Wojciech Samek Marina M.-C. Höhne (née Vidovic)   TU Berlin Fraunhofer HHI BIFOLD Brown University Cornell University OIST Leibniz-Institut für Agrartechnik und Bioökonomie Universität Potsdam

Have had enough of the fake "sources" "cited" by ChatGPT? We have the solution in the form of low-cost causal citations for LLMs: arxiv.org/abs/2505.15807 Thanks to my amazing co-authors Patrick Kahardipraja , Reduan Achtibat, Thomas Wiegand and Wojciech Samek from Fraunhofer HHI

🚀 We’ve released the source code for 𝗔𝗦𝗜𝗗𝗘 (presented as an 𝗢𝗿𝗮𝗹 at the #ICLR2025 BuildTrust workshop)! 🔍ASIDE boosts prompt injection robustness without safety-tuning: we simply rotate embeddings of marked tokens by 90° during instruction-tuning and inference 👇code