AIConfPaper
← Search

Sebastian Lapuschkin

17 accepted papers

2026

ASIDE: Architectural Separation of Instructions and Data in Language Models

Egor Zverev, Evgenii Kortukov, Alexander Panfilov, Alexandra Volkova, Rush Tabesh, Sebastian Lapuschkin +2
ICLR 2026poster

Despite their remarkable performance, large language models lack elementary safety features, making them susceptible to numerous malicious attacks. In particular, previous work has identified the absence of an intrinsic separation between instructions and data as the root cause of the success of pro…

Cited by 0Source ↗code ↗Scholar ↗
2026

Attribution-Guided Decoding

Piotr Komorowski, Elena Golimblevskaia, Reduan Achtibat, Thomas Wiegand, Sebastian Lapuschkin, Wojciech Samek
ICLR 2026poster

The capacity of Large Language Models (LLMs) to follow complex instructions and generate factually accurate text is critical for their real-world application. However, standard decoding methods often fail to robustly satisfy these requirements, while existing control techniques frequently degrade ge…

Cited by 0Source ↗code ↗Scholar ↗
2026

Circuit Insights: Towards Interpretability Beyond Activations

Elena Golimblevskaia, Aakriti Jain, Bruno Puri, Ammar Ibrahim, Wojciech Samek, Sebastian Lapuschkin
ICLR 2026poster

The fields of explainable AI and mechanistic interpretability aim to uncover the internal structure of neural networks, with circuit discovery as a central tool for understanding model computations. Existing approaches, however, rely on manual inspection and remain limited to toy tasks. Automated in…

Cited by 0Source ↗Scholar ↗
2026

Dyslexify: A Mechanistic Defense Against Typographic Attacks in CLIP

Lorenz Hufe, Constantin Venhoff, Maximilian Dreyer, Erblina Purelku, Sebastian Lapuschkin, Wojciech Samek
ICLR 2026poster

Typographic attacks exploit multi-modal systems by injecting text into images, leading to targeted misclassifications, malicious content generation and even Vision-Language Model jailbreaks. In this work, we analyze how CLIP vision encoders behave under typographic attacks, locating specialized atte…

Cited by 0Source ↗code ↗Scholar ↗
2026

PINNfluence: Interpreting PINNs through Influence Functions

Aleksander Krasowski, Jonas Naujoks, Moritz Weckbecker, Galip Yolcu, Thomas Wiegand, Sebastian Lapuschkin +2
ICML 2026poster

Physics-informed neural networks (PINNs) have emerged as a powerful deep learning approach for solving partial differential equations (PDEs) in the physical sciences, yet their behavior remains largely opaque and is typically understood through failure mode analyses rather than explicit interpretabi…

Cited by 0Source ↗Scholar ↗
2026

Strategic Dishonesty Can Undermine AI Safety Evaluations of Frontier LLMs

Alexander Panfilov, Evgenii Kortukov, Kristina Nikolić, Matthias Bethge, Sebastian Lapuschkin, Wojciech Samek +3
ICLR 2026poster

Large language model (LLM) developers aim for their models to be honest, helpful, and harmless. However, when faced with malicious requests, models are trained to refuse, sacrificing helpfulness. We show that frontier LLMs can develop a preference for \textit{dishonesty} as a new strategy, even when…

Cited by 0Source ↗Scholar ↗
2025

FADE: Why Bad Descriptions Happen to Good Features

Bruno Puri, Aakriti Jain, Elena Golimblevskaia, Patrick Kahardipraja, Thomas Wiegand, Wojciech Samek +1
ACL 2025finding

Recent advances in mechanistic interpretability have highlighted the potential of automating interpretability pipelines in analyzing the latent representations within LLMs. While this may enhance our understanding of internal mechanisms, the field lacks standardized evaluation methods for assessing…

Cited by 0Source ↗PDF ↗code ↗Scholar ↗
2025

Manipulating Feature Visualizations with Gradient Slingshots

Dilyara Bareeva, Marina MC Höhne, Alexander Warnecke, Lukas Pirch, Klaus Robert Muller, Konrad Rieck +2
NeurIPS 2025poster

Feature Visualization (FV) is a widely used technique for interpreting concepts learned by Deep Neural Networks (DNNs), which synthesizes input patterns that maximally activate a given feature. Despite its popularity, the trustworthiness of FV explanations has received limited attention. We introduc…

Cited by 0Source ↗code ↗Scholar ↗
2025

Navigating Neural Space: Revisiting Concept Activation Vectors to Overcome Directional Divergence

Frederik Pahde, Maximilian Dreyer, Moritz Weckbecker, Leander Weber, Christopher J. Anders, Thomas Wiegand +2
ICLR 2025poster

With a growing interest in understanding neural network prediction strategies, Concept Activation Vectors (CAVs) have emerged as a popular tool for modeling human-understandable concepts in the latent space. Commonly, CAVs are computed by leveraging linear classifiers optimizing the *separability* o…

Cited by 5Source ↗PDF ↗Scholar ↗
2025

The Atlas of In-Context Learning: How Attention Heads Shape In-Context Retrieval Augmentation

Patrick Kahardipraja, Reduan Achtibat, Thomas Wiegand, Wojciech Samek, Sebastian Lapuschkin
NeurIPS 2025poster

Large language models are able to exploit in-context learning to access external knowledge beyond their training data through retrieval-augmentation. While promising, its inner workings remain unclear. In this work, we shed light on the mechanism of in-context retrieval augmentation for question ans…

Cited by 0Source ↗code ↗Scholar ↗
2024

AttnLRP: Attention-Aware Layer-Wise Relevance Propagation for Transformers

Reduan Achtibat, Sayed Mohammad Vakilzadeh Hatefi, Maximilian Dreyer, Aakriti Jain, Thomas Wiegand, Sebastian Lapuschkin +1
ICML 2024poster

Large Language Models are prone to biased predictions and hallucinations, underlining the paramount importance of understanding their model-internal reasoning process. However, achieving faithful attributions for the entirety of a black-box transformer model and maintaining computational efficiency…

Cited by 32Source ↗PDF ↗code ↗Scholar ↗
2024

CoSy: Evaluating Textual Explanations of Neurons

Laura Kopf, Philine Lou Bommer, Anna Hedström, Sebastian Lapuschkin, Marina MC Höhne, Kirill Bykov
NeurIPS 2024poster

A crucial aspect of understanding the complex nature of Deep Neural Networks (DNNs) is the ability to explain learned concepts within their latent representations. While methods exist to connect neurons to human-understandable textual descriptions, evaluating the quality of these explanations is cha…

Cited by 6Source ↗PDF ↗code ↗Scholar ↗
2024

From Hope to Safety: Unlearning Biases of Deep Models via Gradient Penalization in Latent Space

Maximilian Dreyer, Frederik Pahde, Christopher J. Anders, Wojciech Samek, Sebastian Lapuschkin
AAAI 2024technical

Deep Neural Networks are prone to learning spurious correlations embedded in the training data, leading to potentially biased predictions. This poses risks when deploying these models for high-stake decision-making, such as in medical applications. Current methods for post-hoc model correction eithe…

Cited by 12Source ↗PDF ↗code ↗Scholar ↗
2024

Generative Fractional Diffusion Models

Gabriel Nobis, Maximilian Springenberg, Marco Aversa, Michael Detzel, Rembert Daems, Roderick Murray-Smith +8
NeurIPS 2024poster

We introduce the first continuous-time score-based generative model that leverages fractional diffusion processes for its underlying dynamics. Although diffusion models have excelled at capturing data distributions, they still suffer from various limitations such as slow convergence, mode-collapse o…

Cited by 4Source ↗PDF ↗code ↗Scholar ↗
2023

Shortcomings of Top-Down Randomization-Based Sanity Checks for Evaluations of Deep Neural Network Explanations

Alexander Binder, Leander Weber, Sebastian Lapuschkin, Grégoire Montavon, Klaus-Robert Müller, Wojciech Samek
CVPR 2023poster

While the evaluation of explanations is an important step towards trustworthy models, it needs to be done carefully, and the employed metrics need to be well-understood. Specifically model randomization testing can be overinterpreted if regarded as a primary criterion for selecting or discarding exp…

Cited by 27Source ↗PDF ↗Scholar ↗
2017

Interpretable human action recognition in compressed domain

Vignesh Srinivasan, Sebastian Lapuschkin, Cornelius Hellge, Klaus-Robert Müller, Wojciech Samek
ICASSP 2017accepted

Compressed domain human action recognition algorithms are extremely efficient, because they only require a partial decoding of the video bit stream. However, the question what exactly makes these algorithms decide for a particular action is still a mystery. In this paper, we present a general method…

Cited by 0Source ↗Scholar ↗
2016

Analyzing Classifiers: Fisher Vectors and Deep Neural Networks

Sebastian Lapuschkin, Alexander Binder, Gregoire Montavon, Klaus-Robert Muller, Wojciech Samek
CVPR 2016poster

Fisher vector (FV) classifiers and Deep Neural Networks (DNNs) are popular and successful algorithms for solving image classification problems. However, both are generally considered `black box' predictors as the non-linear transformations involved have so far prevented transparent and interpretable…

Cited by 262PDF ↗Scholar ↗