11 citations · 47 across the 48 of their papers we have counts for
13 papers · 2 filters
Atlas-Alignment: Making Interpretability Transferable Across Language Models
Bruno Puri, Jim Berend, Sebastian Lapuschkin +1
Interpretability is crucial for building safe, reliable, and controllable language models, yet existing interpretability pipelines remain costly and difficult to scale. Interpretin…
LieSolver: PDE-Constrained Learning for IBVPs via Lie Symmetries
René P. Klausen, Ivan Timofeev, Jonas Naujoks +4
Initial-boundary value problems (IBVPs) provide the essential framework for modelling a wide range of phenomena in physics and engineering. We introduce a novel method for efficien…
Circuit Insights: Towards Interpretability Beyond Activations
Elena Golimblevskaia, Aakriti Jain, Bruno Puri +3
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 m…
Strategic Dishonesty Can Undermine AI Safety Evaluations of Frontier LLMs
Alexander Panfilov, Evgenii Kortukov, Kristina Nikolić +6
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, sacrifici…
Attribution-Guided Decoding
Piotr Komorowski, Elena Golimblevskaia, Reduan Achtibat +3
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 dec…
The Algorithm Is Not the Behavior: Learned Priors Override Look-Ahead in a Chess-Playing Neural Network
Elias Sandmann, Sebastian Lapuschkin, Wojciech Samek
Recent mechanistic work has uncovered learned algorithms within neural networks, from modular arithmetic to search and planning in game-playing agents. But does algorithmic structu…
Leveraging Influence Functions for Resampling Data in Physics-Informed Neural Networks
Jonas R. Naujoks, Aleksander Krasowski, Moritz Weckbecker +5
Physics-informed neural networks (PINNs) offer a powerful approach to solving partial differential equations (PDEs), which are ubiquitous in the quantitative sciences. Applied to b…
Attribution-Guided Pruning for Insight and Control: Circuit Discovery and Targeted Correction in Small-scale LLMs
Sayed Mohammad Vakilzadeh Hatefi, Maximilian Dreyer, Reduan Achtibat +5
Large Language Models (LLMs) are widely deployed in real-world applications, yet their internal mechanisms remain difficult to interpret and control, limiting our ability to diagno…
Relevance-driven Input Dropout: an Explanation-guided Regularization Technique
Shreyas Gururaj, Lars Grüne, Wojciech Samek +2
Overfitting is a well-known issue extending even to state-of-the-art (SOTA) Machine Learning (ML) models, resulting in reduced generalization, and a significant train-test performa…
From What to How: Attributing CLIP's Latent Components Reveals Unexpected Semantic Reliance
Maximilian Dreyer, Lorenz Hufe, Jim Berend +3
Transformer-based CLIP models are widely used for text-image probing and feature extraction, making it relevant to understand the internal mechanisms behind their predictions. Whil…
ASIDE: Architectural Separation of Instructions and Data in Language Models
Egor Zverev, Evgenii Kortukov, Alexander Panfilov +5
Despite their remarkable performance, large language models lack elementary safety features, making them susceptible to numerous malicious attacks. In particular, previous work has…
FADE: Why Bad Descriptions Happen to Good Features
Bruno Puri, Aakriti Jain, Elena Golimblevskaia +4
Recent advances in mechanistic interpretability have highlighted the potential of automating interpretability pipelines in analyzing the latent representations within LLMs. While t…
Mechanistic understanding and validation of large AI models with SemanticLens
Maximilian Dreyer, Jim Berend, Tobias Labarta +4
Unlike human-engineered systems such as aeroplanes, where each component's role and dependencies are well understood, the inner workings of AI models remain largely opaque, hinderi…