Prompt engineering for large language models is challenging, as even small prompt perturbations or model changes can significantly impact the generated output texts.Existing evaluation methods of LLM outputs, either automated metrics or human evaluation, have limitations, such as providing limited insights or being labor-intensive.We propose Spotlight, a new approach that combines both automation and human analysis.Based on data mining techniques, we automatically distinguish between random (decoding) variations and systematic differences in language model outputs.This process provides token patterns that describe the systematic differences and guide the user in manually analyzing the effects of their prompts and changes in models efficiently.We create three benchmarks to quantitatively test the reliability of token pattern extraction methods and demonstrate that our approach provides new insights into established prompt data.From a human-centric perspective, through demonstration studies and a user study, we show that our token pattern approach helps users understand the systematic differences of language model outputs.We are further able to discover relevant differences caused by prompt and model changes (e.g.related to gender or culture), thus supporting the prompt engineering process and human-centric model behavior research.

Neural networks are widely adopted to solve complex and challenging tasks. Especially in high-stakes decision-making, understanding their reasoning process is crucial, yet proves challenging for modern deep networks. Feature visualization (FV) is a powerful tool to decode what information neurons are responding to and hence to better understand the reasoning behind such networks. In particular, in FV we generate human-understandable images that reflect the information detected by neurons of interest. However, current methods often yield unrecognizable visualizations, exhibiting repetitive patterns and visual artifacts that are hard to understand for a human. To address these problems, we propose to guide FV through statistics of real image features combined with measures of relevant network flow to generate prototypical images. Our approach yields human-understandable visualizations that both qualitatively and quantitatively improve over state-of-the-art FVs across various architectures. As such, it can be used to decode which information the network uses, complementing mechanistic circuits that identify where it is encoded. Code is available at: https://github.com/adagorgun/VITAL

Deep networks have shown remarkable performance across a wide range of tasks, yet getting a global concept-level understanding of how they function remains a key challenge. Many post-hoc concept-based approaches have been introduced to understand their workings, yet they are not always faithful to the model. Further, they make restrictive assumptions on the concepts a model learns, such as class-specificity, small spatial extent, or alignment to human expectations. In this work, we put emphasis on the faithfulness of such concept-based explanations and propose a new model with model-inherent mechanistic concept-explanations. Our concepts are shared across classes and, from any layer, their contribution to the logit and their input-visualization can be faithfully traced. We also leverage foundation models to propose a new concept-consistency metric, C$^2$-Score, that can be used to evaluate concept-based methods. We show that, compared to prior work, our concepts are quantitatively more consistent and users find our concepts to be more interpretable, all while retaining competitive ImageNet performance.

Our sense of touch plays a crucial role in physical collaboration, yet rendering realistic haptic feedback in collaborative extended reality (XR) remains a challenge. Co-located XR systems predominantly rely on prefabricated passive props that provide high-fidelity interaction but offer limited adaptability. Haptic Illusions (HIs), which leverage multisensory integration, have proven effective in expanding haptic experiences in single-user contexts. However, their role in XR collaboration has not been explored. To examine the applicability of HIs in multi-user scenarios, we conducted an experimental user study (N=30) investigating their effect on a collaborative object handover task in virtual reality. We manipulated visual shape and size individually and analyzed their impact on users' performance, experience, and behavior. Results show that while participants adapted to the illusions by shifting sensory reliance and employing specific sensorimotor strategies, visuo-haptic mismatches reduced both performance and experience. Moreover, mismatched visualizations in asymmetric user roles negatively impacted performance. Drawing from these findings, we provide practical guidelines for incorporating HIs into collaborative XR, marking a first step toward richer haptic interactions in shared virtual spaces.

Discovering patterns in data that best describe the differences between classes allows to hypothesize and reason about class-specific mechanisms. In molecular biology, for example, these bear the promise of advancing the understanding of cellular processes differing between tissues or diseases, which could lead to novel treatments. To be useful in practice, methods that tackle the problem of finding such differential patterns have to be readily interpretable by domain experts, and scalable to the extremely high-dimensional data. In this work, we propose a novel, inherently interpretable binary neural network architecture Diffnaps that extracts differential patterns from data. Diffnaps is scalable to hundreds of thousands of features and robust to noise, thus overcoming the limitations of current state-of-the-art methods in large-scale applications such as in biology. We show on synthetic and real world data, including three biological applications, that unlike its competitors, Diffnaps consistently yields accurate, succinct, and interpretable class descriptions.

The practical utility of machine learning models in the sciences often hinges on their interpretability. It is common to assess a model's merit for scientific discovery, and thus novel insights, by how well it aligns with already available domain knowledge - a dimension that is currently largely disregarded in the comparison of neural network models. While pruning can simplify deep neural network architectures and excels in identifying sparse models, as we show in the context of gene regulatory network inference, state-of-the-art techniques struggle with biologically meaningful structure learning. To address this issue, we propose DASH, a generalizable framework that guides network pruning by using domain-specific structural information in model fitting and leads to sparser, better interpretable models that are more robust to noise. Using both synthetic data with ground truth information, as well as real-world gene expression data, we show that DASH, using knowledge about gene interaction partners within the putative regulatory network, outperforms general pruning methods by a large margin and yields deeper insights into the biological systems being studied.

MOTIVATION: DNA CpG methylation (CpGm) has proven to be a crucial epigenetic factor in the mammalian gene regulatory system. Assessment of DNA CpG methylation values via whole-genome bisulfite sequencing (WGBS) is, however, computationally extremely demanding. RESULTS: We present FAst MEthylation calling (FAME), the first approach to quantify CpGm values directly from bulk or single-cell WGBS reads without intermediate output files. FAME is very fast but as accurate as standard methods, which first produce BS alignment files before computing CpGm values. We present experiments on bulk and single-cell bisulfite datasets in which we show that data analysis can be significantly sped-up and help addressing the current WGBS analysis bottleneck for large-scale datasets without compromising accuracy. AVAILABILITY AND IMPLEMENTATION: An implementation of FAME is open source and licensed under GPL-3.0 at https://github.com/FischerJo/FAME.

State-of-the-art deep learning methods achieve human-like performance on many tasks, but make errors nevertheless. Characterizing these errors in easily interpretable terms gives insight into whether a classifier is prone to making systematic errors, but also gives a way to act and improve the classifier. We propose to discover those feature-value combinations (i.e., patterns) that strongly correlate with correct resp. erroneous predictions to obtain a global and interpretable description for arbitrary classifiers. We show this is an instance of the more general label description problem, which we formulate in terms of the Minimum Description Length principle. To discover a good pattern set, we develop the efficient Premise algorithm. Through an extensive set of experiments we show it performs very well in practice on both synthetic and real-world data. Unlike existing solutions, it ably recovers ground truth patterns, even on highly imbalanced data over many features. Through two case studies on Visual Question Answering and Named Entity Recognition, we confirm that Premise gives clear and actionable insight into the systematic errors made by modern NLP classifiers.

Estimating mutual information (MI) between two continuous random variables X and Y allows to capture non-linear dependencies between them, non-parametrically. As such, MI estimation lies at the core of many data science applications. Yet, robustly estimating MI for high-dimensional X and Y is still an open research question. In this paper, we formulate this problem through the lens of manifold learning. That is, we leverage the common assumption that the information of X and Y is captured by a low-dimensional manifold embedded in the observed high-dimensional space and transfer it to MI estimation. As an extension to state-of-the-art kNN estimators, we propose to determine the k-nearest neighbors via geodesic distances on this manifold rather than from the ambient space, which allows us to estimate MI even in the high-dimensional setting. An empirical evaluation of our method, G-KSG, against the state-of-the-art shows that it yields good estimations of MI in classical benchmark and manifold tasks, even for high dimensional datasets, which none of the existing methods can provide.

We propose a novel method for exploring how neurons within neural networks interact. In particular, we consider activation values of a network for given data, and propose to mine noise-robust rules of the form X {\rightarrow} Y , where X and Y are sets of neurons in different layers. We identify the best set of rules by the Minimum Description Length Principle as the rules that together are most descriptive of the activation data. To learn good rule sets in practice, we propose the unsupervised ExplaiNN algorithm. Extensive evaluation shows that the patterns it discovers give clear insight in how networks perceive the world: they identify shared, respectively class-specific traits, compositionality within the network, as well as locality in convolutional layers. Moreover, these patterns are not only easily interpretable, but also supercharge prototyping as they identify which groups of neurons to consider in unison.

Pattern set mining has been successful in discovering small sets of highly informative and useful patterns from data. To find good models, existing methods heuristically explore the twice-exponential search space over all possible pattern sets in a combinatorial way, by which they are limited to data over at most hundreds of features, as well as likely to get stuck in local minima. Here, we propose a gradient based optimization approach that allows us to efficiently discover high-quality pattern sets from data of millions of rows and hundreds of thousands of features.

Pattern mining is one of the core topics of data mining. We consider the problem of mining a succinct set of patterns that together explain the data in terms of mutual exclusivity and co-occurence. That is, we extend the traditional pattern languages beyond conjunctions, enabling us to capture more complex relationships, such as replacable sub-components or antagonists in biological pathways.

Multi-pass stereoscopic video rendering (MSVR) allows to present stereoscopic video streams within game engines on consumer graphics boards. We present our approach and discuss aspects of performance and occlusion of virtual objects.