Publications

Thaddäus Wiedemer, Yash Sharma, Ameya Prabhu, matthias-bethge, Wieland Brendel (2025). Pretraining Frequency Predicts Compositional Generalization of CLIP on Real-World Tasks.

Prasanna Mayilvahanan, Thaddäus Wiedemer, Sayak Mallick, matthias-bethge, Wieland Brendel (2025). LLMs on the Line: Data Determines Loss-to-Loss Scaling Laws.

Jack Brady, Julius von Kügelgen, Sébastien Lachapelle, Simon Buchholz, Thomas Kipf, Wieland Brendel (2024). Interaction Asymmetry: A General Principle for Learning Composable Abstractions.

Patrik Reizinger, Alice Bizeul, Attila Juhos, Julia E Vogt, Randall Balestriero, Wieland Brendel, David Klindt (2024). Cross-Entropy Is All You Need To Invert the Data Generating Process.

Prasanna Mayilvahanan, Roland S Zimmermann, Thaddäus Wiedemer, Evgenia Rusak, Attila Juhos, matthias-bethge, Wieland Brendel (2024). In search of forgotten domain generalization.

Thomas Klein, Wieland Brendel, Felix Wichmann (2024). Error consistency between humans and machines as a function of presentation duration.

Anna Mészáros, Szilvia Ujváry, Wieland Brendel, Patrik Reizinger, Ferenc Huszár (2024). Rule Extrapolation in Language Models: A Study of Compositional Generalization on OOD Prompts.

Evgenia Rusak, Patrik Reizinger, Attila Juhos, Oliver Bringmann, Roland S Zimmermann, Wieland Brendel (2024). InfoNCE: Identifying the Gap Between Theory and Practice.

Josue O Caro, Yilong Ju, Ryan Pyle, Sourav Dey, Wieland Brendel, Fabio Anselmi, Ankit B Patel (2024). Translational symmetry in convolutions with localized kernels causes an implicit bias toward high frequency adversarial examples.

Patrik Reizinger, Siyuan Guo, Ferenc Huszár, Bernhard Schölkopf, Wieland Brendel (2024). Identifiable Exchangeable Mechanisms for Causal Structure and Representation Learning.

Patrik Reizinger, Szilvia Ujváry, Anna Mészáros, Anna Kerekes, Wieland Brendel, Ferenc Huszár (2024). Understanding LLMs Requires More Than Statistical Generalization.

Patrik Reizinger, Szilvia Ujváry, Anna Mészáros, Anna Kerekes, Wieland Brendel, Ferenc Huszár (2024). Position: Understanding LLMs requires more than statistical generalization.

Goutham Rajendran, Patrik Reizinger, Wieland Brendel, Pradeep Kumar Ravikumar (2024). An interventional perspective on identifiability in gaussian lti systems with independent component analysis.

Amro Abbas, Evgenia Rusak, Kushal Tirumala, Wieland Brendel, Kamalika Chaudhuri, Ari S Morcos (2024). Effective pruning of web-scale datasets based on complexity of concept clusters.

Prasanna Mayilvahanan, Thaddäus Wiedemer, Evgenia Rusak, matthias-bethge, Wieland Brendel (2023). Does CLIP's Generalization Performance Mainly Stem from High Train-Test Similarity?.

Thaddäus Wiedemer, Jack Brady, Alexander Panfilov, Attila Juhos, matthias-bethge, Wieland Brendel (2023). Provable Compositional Generalization for Object-Centric Learning.

Thomas Klein, Wieland Brendel, Felix Wichmann (2023). Feature Visualizations do not sufficiently explain hidden units of Artificial Neural Networks.

Roland S Zimmermann, Thomas Klein, Wieland Brendel (2023). Scale Alone Does not Improve Mechanistic Interpretability in Vision Models.

Thaddäus Wiedemer, Prasanna Mayilvahanan, matthias-bethge, Wieland Brendel (2023). Compositional Generalization from First Principles.

Robert Geirhos, Roland S Zimmermann, Blair Bilodeau, Wieland Brendel, Been Kim (2023). Don't trust your eyes: on the (un) reliability of feature visualizations.

Jack Brady, Roland S Zimmermann, Yash Sharma, Bernhard Schölkopf, Julius von Kügelgen, Wieland Brendel (2023). Provably Learning Object-Centric Representations.

Zhe Li, Josue Ortega Caro, Evgenia Rusak, Wieland Brendel, matthias-bethge, Fabio Anselmi, Ankit B Patel, Andreas S Tolias, Xaq Pitkow (2023). Robust deep learning object recognition models rely on low frequency information in natural images.

Max Wolff, Wieland Brendel, Stuart Wolff (2023). The Independent Compositional Subspace Hypothesis for the Structure of CLIP's Last Layer.

Zhe Li, Wieland Brendel, Edgar Y Walker, Erick Cobos, Taliah Muhammad, Jacob Reimer, matthias-bethge, Fabian H Sinz, Xaq Pitkow, Andreas S Tolias (2023). Learning From Brains How to Regularize Machines (Supplementary Material).

Patrik Reizinger, Yash Sharma, matthias-bethge, Bernhard Schölkopf, Ferenc Huszár, Wieland Brendel (2023). Jacobian-based Causal Discovery with Nonlinear ICA.

Dejana Mandić, Wieland Brendel, Claudio Michaelis (2023). Iterative weakly supervised learning for novel class object detection.

Robert Geirhos, Kantharaju Narayanappa, Benjamin Mitzkus, Tizian Thieringer, matthias-bethge, Felix A Wichmann, Wieland Brendel (2022). The bittersweet lesson: data-rich models narrow the behavioural gap to human vision.

Evgenia Rusak, Steffen Schneider, George Pachitariu, Luisa Eck, Peter Vincent Gehler, Oliver Bringmann, Wieland Brendel, matthias-bethge (2022). If your data distribution shifts, use self-learning.

Roland S Zimmermann, Wieland Brendel, Florian Tramer, Nicholas Carlini (2022). Increasing Confidence in Adversarial Robustness Evaluations.

Patrik Reizinger, Luigi Gresele, Jack Brady, Julius von Kügelgen, Dominik Zietlow, Bernhard Schölkopf, Georg Martius, Wieland Brendel, Michel Besserve (2022). Embrace the Gap: VAEs Perform Independent Mechanism Analysis.

Evgenia Rusak, Steffen Schneider, Peter Vincent Gehler, Oliver Bringmann, Wieland Brendel, matthias-bethge (2022). Imagenet-d: A new challenging robustness dataset inspired by domain adaptation.

Evgenia Rusak, Patrik Reizinger, Roland S Zimmermann, Oliver Bringmann, Wieland Brendel (2022). Content suppresses style: dimensionality collapse in contrastive learning.

Wieland Brendel, Jonas Rauber, matthias-bethge (2021). 2.1 Decision-Based Adversarial Attacks.

Lukas Schott, Julius von Kügelgen, Frederik Träuble, Peter Gehler, Chris Russell, matthias-bethge, Bernhard Schölkopf, Francesco Locatello, Wieland Brendel (2021). Visual representation learning does not generalize strongly within the same domain.

Roland S Zimmermann, Judy Borowski, Robert Geirhos, matthias-bethge, Thomas SA Wallis, Wieland Brendel (2021). How Well do Feature Visualizations Support Causal Understanding of CNN Activations?.

Robert Geirhos, Kantharaju Narayanappa, Benjamin Mitzkus, Tizian Thieringer, matthias-bethge, Felix A Wichmann, Wieland Brendel (2021). Partial success in closing the gap between human and machine vision.

Julius von Kügelgen, Yash Sharma, Luigi Gresele, Wieland Brendel, Bernhard Schölkopf, Michel Besserve, Francesco Locatello (2021). Self-Supervised Learning with Data Augmentations Provably Isolates Content from Style.

Evgenia Rusak, Steffen Schneider, Peter Gehler, Oliver Bringmann, Wieland Brendel, matthias-bethge (2021). Adapting imagenet-scale models to complex distribution shifts with self-learning.

Christina M Funke, Judy Borowski, Karolina Stosio, Wieland Brendel, Thomas SA Wallis, matthias-bethge (2021). Five points to check when comparing visual perception in humans and machines.

Maura Pintor, Fabio Roli, Wieland Brendel, Battista Biggio (2021). Fast Minimum-norm Adversarial Attacks through Adaptive Norm Constraints.

Roland S Zimmermann, Yash Sharma, Steffen Schneider, matthias-bethge, Wieland Brendel (2021). Contrastive Learning Inverts the Data Generating Process.

CM Funke, J Borowski, K Stosio, W Brendel, TSA Wallis, M Bethge (2021). Five points to check when com.

Judy Borowski, Roland S Zimmermann, Judith Schepers, Robert Geirhos, Thomas SA Wallis, matthias-bethge, Wieland Brendel (2020). Exemplary Natural Images Explain CNN Activations Better than Feature Visualizations.

Robert Geirhos, Jörn-Henrik Jacobsen, Claudio Michaelis, Richard Zemel, Wieland Brendel, matthias-bethge, Felix A Wichmann (2020). Unintended cue learning: Lessons for deep learning from experimental psychology.

Robert Geirhos, Kantharaju Narayanappa, Benjamin Mitzkus, matthias-bethge, Felix A Wichmann, Wieland Brendel (2020). On the surprising similarities between supervised and self-supervised models.

Jonas Rauber, Roland Zimmermann, matthias-bethge, Wieland Brendel (2020). Foolbox native: Fast adversarial attacks to benchmark the robustness of machine learning models in pytorch, tensorflow, and jax.

Jonas Rauber, matthias-bethge, Wieland Brendel (2020). EagerPy: Writing code that works natively with PyTorch, TensorFlow, JAX, and NumPy.

David Klindt, Lukas Schott, Yash Sharma, Ivan Ustyuzhaninov, Wieland Brendel, matthias-bethge, Dylan Paiton (2020). Towards Nonlinear Disentanglement in Natural Data with Temporal Sparse Coding.

Steffen Schneider, Evgenia Rusak, Luisa Eck, Oliver Bringmann, Wieland Brendel, matthias-bethge (2020). Improving robustness against common corruptions by covariate shift adaptation.

Josue Ortega Caro, Yilong Ju, Ryan Pyle, Sourav Dey, Wieland Brendel, Fabio Anselmi, Ankit Patel (2020). Local convolutions cause an implicit bias towards high frequency adversarial examples.

Marissa A Weis, Kashyap Chitta, Yash Sharma, Wieland Brendel, matthias-bethge, Andreas Geiger, Alexander S Ecker (2020). Unmasking the inductive biases of unsupervised object representations for video sequences.

Josue Ortega Caro, Yilong Ju, Ryan Pyle, Sourav Dey, Wieland Brendel, Fabio Anselmi, Ankit Patel (2020). Local Convolutions Cause an Implicit Bias towards High Frequency Adversarial Examples.

Robert Geirhos, Jörn-Henrik Jacobsen, Claudio Michaelis, Richard Zemel, Wieland Brendel, matthias-bethge, Felix A Wichmann (2020). Shortcut Learning in Deep Neural Networks.

Wieland Brendel, Ralph Bourdoukan, Pietro Vertechi, Christian K Machens, Sophie Denéve (2020). Learning to represent signals spike by spike.

Florian Tramer, Nicholas Carlini, Wieland Brendel, Aleksander Madry (2020). On adaptive attacks to adversarial example defenses.

Evgenia Rusak, Lukas Schott, Roland Zimmermann, Julian Bitterwolf, Oliver Bringmann, matthias-bethge, Wieland Brendel (2020). Increasing the robustness of DNNs against image corruptions by playing the Game of Noise.

Steffen Schneider123, Shubham Krishna, Luisa Eck, Wieland Brendel, Mackenzie W Mathis, matthias-bethge (2020). Generalized Invariant Risk Minimization: relating adaptation and invariant representation learning.

Evgenia Rusak, Lukas Schott, Roland S Zimmermann, Julian Bitterwolf, Oliver Bringmann, matthias-bethge, Wieland Brendel (2020). A simple way to make neural networks robust against diverse image corruptions.

Robert Geirhos, Patricia Rubisch, Jonas Rauber, Carlos R Medina Temme, Claudio Michaelis, Wieland Brendel, matthias-bethge, Felix A Wichmann (2019). Inducing a human-like shape bias leads to emergent human-level distortion robustness in CNNs.

Claudio Michaelis, Benjamin Mitzkus, Robert Geirhos, Evgenia Rusak, Oliver Bringmann, Alexander S Ecker, matthias-bethge, Wieland Brendel (2019). Benchmarking robustness in object detection: Autonomous driving when winter is coming.

Judith Borowski, Wieland Brendel, matthias-bethge (2019). Reproducing Decision-Making With Constrained Networks to Understand Deep Neural Networks.

Wieland Brendel, matthias-bethge (2019). Approximating cnns with bag-of-local-features models works surprisingly well on imagenet.

Nicholas Carlini, Anish Athalye, Nicolas Papernot, Wieland Brendel, Jonas Rauber, Dimitris Tsipras, Ian Goodfellow, Aleksander Madry, Alexey Kurakin (2019). On evaluating adversarial robustness.

Zhe Li, Wieland Brendel, Edgar Walker, Erick Cobos, Taliah Muhammad, Jacob Reimer, matthias-bethge, Fabian Sinz, Zachary Pitkow, Andreas Tolias (2019). Learning from brains how to regularize machines.

Wieland Brendel, Jonas Rauber, Matthias Kümmerer, Ivan Ustyuzhaninov, matthias-bethge (2019). Accurate, reliable and fast robustness evaluation.

Robert Geirhos, Patricia Rubisch, Claudio Michaelis, matthias-bethge, Felix A Wichmann, Wieland Brendel (2018). ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness.

Christina Funke, Judy Borowski, Thomas Wallis, Wieland Brendel, Alexander Ecker, matthias-bethge (2018). Comparing the ability of humans and DNNs to recognise closed contours in cluttered images.

Maysam Oladazimi, Wieland Brendel, Cornelius Schwarz (2018). Biomechanical texture coding in rat whiskers.

Ivan Ustyuzhaninov, Claudio Michaelis, Wieland Brendel, matthias-bethge (2018). One-shot texture segmentation.

Lukas Schott, Jonas Rauber, matthias-bethge, Wieland Brendel (2018). Towards the first adversarially robust neural network model on MNIST.

Alexander Böttcher, Wieland Brendel, Bernhard Englitz, matthias-bethge (2018). Trace your sources in large-scale data: one ring to find them all.

Wieland Brendel, Jonas Rauber, matthias-bethge (2017). Decision-based adversarial attacks: Reliable attacks against black-box machine learning models.

Jonas Rauber, Wieland Brendel, matthias-bethge (2017). Foolbox v0. 8.0: A python toolbox to benchmark the robustness of machine learning models.

Wieland Brendel, matthias-bethge (2017). Comment on" Biologically inspired protection of deep networks from adversarial attacks".

Ivan Ustyuzhaninov, Wieland Brendel, Leon A Gatys, matthias-bethge (2017). What does it take to generate natural textures?.

A Boettcher, W Brendel, M Bethge (2016). Large scale blind source separation.

I Ustyuzhaninov, W Brendel, L Gatys, M Bethge (2016). Texture synthesis using random shallow neural networks.

M Bethge, W Brendel (2016). Deep Neural Networks: A New Tool for Understanding the Brain.

Ivan Ustyuzhaninov, Wieland Brendel, Leon A Gatys, matthias-bethge (2016). Texture synthesis using shallow convolutional networks with random filters.

Dmitry Kobak, Wieland Brendel, Christos Constantinidis, Claudia E Feierstein, Adam Kepecs, Zachary F Mainen, Xue-Lian Qi, Ranulfo Romo, Naoshige Uchida, Christian K Machens (2016). Demixed principal component analysis of neural population data.

Wieland Brendel (2014). On the demixing and self-organized formation of neural population responses.

Pietro Vertechi, Wieland Brendel, Christian K Machens (2014). Unsupervised learning of an efficient short-term memory network.

Wieland Brendel, Ranulfo Romo, Christian K Machens (2011). Demixed principal component analysis.

Wieland Brendel, Michael Thies (2010). Covariant boost and structure functions of baryons in Gross-Neveu models.

Wieland Brendel, Falk Bruckmann, Lukas Janssen, Andreas Wipf, Christian Wozar (2009). Instanton constituents and fermionic zero modes in twisted CPn models.