arXiv subjects

Przemysław Biecek

Publications and source records attributed to Przemysław Biecek.

At least 19 recordsLinked to original sources

arXiv2026

Modelpedia: A Catalog of Model Findings for the Meta-Science of AI

Scientific knowledge about AI models is produced faster than the community can organize it. Every few months a new foundation model reshapes the field and hundreds of papers, blogs, and technical reports document how each behaves or fails. Yet, these findings remain scattered and effectively unretrievable. To address this gap we present Modelpedia, an automated, LLM-assisted framework that extracts findings about models from published papers, links it to the model, dataset, method, and concept it concerns, and aggregates the result into a searchable public catalog. Applying the prototype to accepted ICLR 2024 and 2025 papers, we extract over a thousand findings and, treating the catalog itself as an object of study, run a meta-analysis of how the community investigates models. Now, we invite the community to explore, contribute to, and build on the open catalog, and to help establish model findings as a shared foundation for the meta-science of AI.

cs.LG↗

arXiv2026

Interpretability-Guided Soft Pruning of Attention Heads in Vision Transformers

Vision foundation models, such as DINOv2, learn highly expressive representations but rely on massive, opaque architectures that demand substantial computational power and memory. To provide an interpretable-guided and efficient solution to this issue, we first propose a spectral analysis and new visualization technique for individual attention heads based on the Laplacian eigenvectors of their attention maps. Building upon recent observations regarding the block structure of Vision Transformers, we perform semantic clustering of attention heads and identify functional redundancies. Leveraging these insights, we introduce SAPER (Soft Attention PrunER), an end-to-end differentiable pruning framework based on the LapSum Soft Top-K approach. Extensive experiments on ImageNet-1K demonstrate that SAPER achieves a highly favorable accuracy-efficiency trade-off, outperforming the competitive RAPTOR baseline in FLOPs reduction while preserving strong classification performance.

cs.CV↗

arXiv2026

Explaining BiomedCLIP with Weighted Banzhaf Interactions Supported by Tree-Gram Parsing

Vision-Language Models (VLMs) are demonstrating significant capabilities in medical tasks like radiology analysis, yet providing faithful and interpretable explanations remains a key consideration for their responsible deployment in clinical settings. However, existing explanation methods, such as the widely used FIxLIP framework, often struggle with the fine-grained nature of modern tokenizers. The tokenization problem fragments clinical concepts---splitting terms like "saddle embolus" into scattered, meaningless subwords---which leads to noisy, semantically incoherent cross-modal attributions. Such fragmentation also results in a combinatorial explosion of interaction possibilities, obscuring the model's true reasoning. To address this, we introduce ParseFIxLIP, an extension that incorporates the Tree-Gram Parsing into the Banzhaf interaction game used by FIxLIP. This semantically informed strategy utilizes dependency parsing trees to define explanation players by grouping related text tokens into semantically coherent units. Our smart_depth grouping strategy, merging tokens according to spaCy token dependency tree, successfully mitigates concept fragmentation, yielding substantially more interpretable cross-modal interactions by unifying complex medical concepts. Quantitatively, while baselines struggled with the high dimensionality of long captions, our parsing approach maintained statistical robustness and semantic parsimony. Qualitative analysis on BiomedCLIP, validated on medical imagery (ROCOv2) and general examples, confirms that the approach accurately captures the synergistic influence of grouped words on model predictions. In conclusion, our work offers intuitive and clinically relevant insights into VLM decision-making, fulfilling the critical need for coherent explanations in the medical domain.

cs.CV↗

arXiv2026

Your CLIP has 164 dimensions of noise: Exploring the embeddings covariance eigenspectrum of contrastively pretrained vision-language transformers

Contrastively pre-trained Vision-Language Models (VLMs) serve as powerful feature extractors. Yet, their shared latent spaces are prone to structural anomalies and act as repositories for non-semantic, multi-modal noise. To address this phenomenon, we employ spectral decomposition of covariance matrices to decompose the VLM latent space into a multi-modal semantic signal component and a shared noise subspace. We observe that this noise geometry exhibits strong subgroup invariance across distinct data subsets. Crucially, pruning these shared noise dimensions is mainly harmless, preserving or actively improving downstream task performance. By isolating true semantic signals from artifactual noise, this work provides new mechanistic insights into the representational structure of modern VLMs, suggesting that a substantial fraction of their latent geometry is governed by shared, architecture-level noise rather than task-relevant semantics alone.

cs.CV↗

arXiv2026

LINE: LLM-based Iterative Neuron Explanations for Vision Models

Interpreting individual neurons in deep neural networks is a crucial step towards understanding their complex decision-making processes and ensuring AI safety. Despite recent progress in neuron labeling, existing methods often limit the search space to predefined concept vocabularies or produce overly specific descriptions that fail to capture higher-order, global concepts. We introduce LINE, a novel, training-free iterative approach tailored for open-vocabulary concept labeling in vision models. Operating in a strictly black-box setting, LINE leverages a large language model and a text-to-image generator to iteratively propose and refine concepts in a closed loop, guided by activation history. LINE achieves state-of-the-art performance across multiple model architectures, yielding AUC improvements of up to 0.11 on ImageNet and 0.05 on Places365, while discovering, on average, 27% of new concepts missed by predefined vocabularies. Beyond identifying the top concept, LINE provides a complete generation history, enabling polysemanticity evaluation and producing visual explanations that rival gradient-dependent activation maximization methods. The source code will be made available soon.

cs.CV↗

arXiv2026

Trojan horse hunt in deep forecasting models: Insights from the European Space Agency competition

Forecasting plays a crucial role in modern safety-critical applications, such as space operations. However, the increasing use of deep forecasting models introduces a new security risk of trojan horse attacks, carried out by hiding a backdoor in the training data or directly in the model weights. Once implanted, the backdoor is activated by a specific trigger pattern at test time, causing the model to produce manipulated predictions. We focus on this issue in our \textit{Trojan Horse Hunt} data science competition, where more than 200 teams faced the task of identifying triggers hidden in deep forecasting models for spacecraft telemetry. We describe the novel task formulation, benchmark set, evaluation protocol, and best solutions from the competition. We further summarize key insights and research directions for effective identification of triggers in time series forecasting models. All materials are publicly available on the official competition webpage https://www.kaggle.com/competitions/trojan-horse-hunt-in-space.

cs.LG↗

arXiv2026

Analyzing the Temporal Factors for Anxiety and Depression Symptoms with the Rashomon Perspective

This paper introduces a new modeling perspective in the public mental health domain to provide a robust interpretation of the relations between anxiety and depression, and the demographic and temporal factors. This perspective particularly leverages the Rashomon Effect, where multiple models exhibit similar predictive performance but rely on diverse internal structures. Instead of considering these multiple models, choosing a single best model risks masking alternative narratives embedded in the data. To address this, we employed this perspective in the interpretation of a large-scale psychological dataset, specifically focusing on the Patient Health Questionnaire-4. We use a random forest model combined with partial dependence profiles to rigorously assess the robustness and stability of predictive relationships across the resulting Rashomon set, which consists of multiple models that exhibit similar predictive performance. Our findings confirm that demographic variables \texttt{age}, \texttt{sex}, and \texttt{education} lead to consistent structural shifts in anxiety and depression risk. Crucially, we identify significant temporal effects: risk probability demonstrates clear diurnal and circaseptan fluctuations, peaking during early morning hours. This work demonstrates the necessity of moving beyond the best model to analyze the entire Rashomon set. Our results highlight that the observed variability, particularly due to circadian and circaseptan rhythms, must be meticulously considered for robust interpretation in psychological screening. We advocate for a multiplicity-aware approach to enhance the stability and generalizability of ML-based conclusions in mental health research.

stat.AP↗

arXiv2025

Fake or Real: The Impostor Hunt in Texts for Space Operations

The "Fake or Real" competition hosted on Kaggle (https://www.kaggle.com/competitions/fake-or-real-the-impostor-hunt ) is the second part of a series of follow-up competitions and hackathons related to the "Assurance for Space Domain AI Applications" project funded by the European Space Agency (https://assurance-ai.space-codev.org/ ). The competition idea is based on two real-life AI security threats identified within the project -- data poisoning and overreliance in Large Language Models. The task is to distinguish between the proper output from LLM and the output generated under malicious modification of the LLM. As this problem was not extensively researched, participants are required to develop new techniques to address this issue or adjust already existing ones to this problem's statement.

cs.LG↗

arXiv2025

Beyond the Single-Best Model: Rashomon Partial Dependence Profile for Trustworthy Explanations in AutoML

Automated machine learning systems efficiently streamline model selection but often focus on a single best-performing model, overlooking explanation uncertainty, an essential concern in human centered explainable AI. To address this, we propose a novel framework that incorporates model multiplicity into explanation generation by aggregating partial dependence profiles (PDP) from a set of near optimal models, known as the Rashomon set. The resulting Rashomon PDP captures interpretive variability and highlights areas of disagreement, providing users with a richer, uncertainty aware view of feature effects. To evaluate its usefulness, we introduce two quantitative metrics, the coverage rate and the mean width of confidence intervals, to evaluate the consistency between the standard PDP and the proposed Rashomon PDP. Experiments on 35 regression datasets from the OpenML CTR23 benchmark suite show that in most cases, the Rashomon PDP covers less than 70% of the best model's PDP, underscoring the limitations of single model explanations. Our findings suggest that Rashomon PDP improves the reliability and trustworthiness of model interpretations by adding additional information that would otherwise be neglected. This is particularly useful in high stakes domains where transparency and confidence are critical.

cs.LG↗

arXiv2025

Trojan Horse Hunt in Time Series Forecasting for Space Operations

This competition hosted on Kaggle (https://www.kaggle.com/competitions/trojan-horse-hunt-in-space) is the first part of a series of follow-up competitions and hackathons related to the "Assurance for Space Domain AI Applications" project funded by the European Space Agency (https://assurance-ai.space-codev.org/). The competition idea is based on one of the real-life AI security threats identified within the project -- the adversarial poisoning of continuously fine-tuned satellite telemetry forecasting models. The task is to develop methods for finding and reconstructing triggers (trojans) in advanced models for satellite telemetry forecasting used in safety-critical space operations. Participants are provided with 1) a large public dataset of real-life multivariate satellite telemetry (without triggers), 2) a reference model trained on the clean data, 3) a set of poisoned neural hierarchical interpolation (N-HiTS) models for time series forecasting trained on the dataset with injected triggers, and 4) Jupyter notebook with the training pipeline and baseline algorithm (the latter will be published in the last month of the competition). The main task of the competition is to reconstruct a set of 45 triggers (i.e., short multivariate time series segments) injected into the training data of the corresponding set of 45 poisoned models. The exact characteristics (i.e., shape, amplitude, and duration) of these triggers must be identified by participants. The popular Neural Cleanse method is adopted as a baseline, but it is not designed for time series analysis and new approaches are necessary for the task. The impact of the competition is not limited to the space domain, but also to many other safety-critical applications of advanced time series analysis where model poisoning may lead to serious consequences.

cs.LG↗

arXiv2025

MASCOTS: Model-Agnostic Symbolic COunterfactual explanations for Time Series

Counterfactual explanations provide an intuitive way to understand model decisions by identifying minimal changes required to alter an outcome. However, applying counterfactual methods to time series models remains challenging due to temporal dependencies, high dimensionality, and the lack of an intuitive human-interpretable representation. We introduce MASCOTS, a method that leverages the Bag-of-Receptive-Fields representation alongside symbolic transformations inspired by Symbolic Aggregate Approximation. By operating in a symbolic feature space, it enhances interpretability while preserving fidelity to the original data and model. Unlike existing approaches that either depend on model structure or autoencoder-based sampling, MASCOTS directly generates meaningful and diverse counterfactual observations in a model-agnostic manner, operating on both univariate and multivariate data. We evaluate MASCOTS on univariate and multivariate benchmark datasets, demonstrating comparable validity, proximity, and plausibility to state-of-the-art methods, while significantly improving interpretability and sparsity. Its symbolic nature allows for explanations that can be expressed visually, in natural language, or through semantic representations, making counterfactual reasoning more accessible and actionable.

cs.LG↗

arXiv2025

The Role of Hyperparameters in Predictive Multiplicity

This paper investigates the critical role of hyperparameters in predictive multiplicity, where different machine learning models trained on the same dataset yield divergent predictions for identical inputs. These inconsistencies can seriously impact high-stakes decisions such as credit assessments, hiring, and medical diagnoses. Focusing on six widely used models for tabular data - Elastic Net, Decision Tree, k-Nearest Neighbor, Support Vector Machine, Random Forests, and Extreme Gradient Boosting - we explore how hyperparameter tuning influences predictive multiplicity, as expressed by the distribution of prediction discrepancies across benchmark datasets. Key hyperparameters such as lambda in Elastic Net, gamma in Support Vector Machines, and alpha in Extreme Gradient Boosting play a crucial role in shaping predictive multiplicity, often compromising the stability of predictions within specific algorithms. Our experiments on 21 benchmark datasets reveal that tuning these hyperparameters leads to notable performance improvements but also increases prediction discrepancies, with Extreme Gradient Boosting exhibiting the highest discrepancy and substantial prediction instability. This highlights the trade-off between performance optimization and prediction consistency, raising concerns about the risk of arbitrary predictions. These findings provide insight into how hyperparameter optimization leads to predictive multiplicity. While predictive multiplicity allows prioritizing domain-specific objectives such as fairness and reduces reliance on a single model, it also complicates decision-making, potentially leading to arbitrary or unjustified outcomes.

cs.LG↗

arXiv2025

Mind What You Ask For: Emotional and Rational Faces of Persuasion by Large Language Models

Be careful what you ask for, you just might get it. This saying fits with the way large language models (LLMs) are trained, which, instead of being rewarded for correctness, are increasingly rewarded for pleasing the recipient. So, they are increasingly effective at persuading us that their answers are valuable. But what tricks do they use in this persuasion? In this study, we examine what are the psycholinguistic features of the responses used by twelve different language models. By grouping response content according to rational or emotional prompts and exploring social influence principles employed by LLMs, we ask whether and how we can mitigate the risks of LLM-driven mass misinformation. We position this study within the broader discourse on human-centred AI, emphasizing the need for interdisciplinary approaches to mitigate cognitive and societal risks posed by persuasive AI responses.

cs.CL↗

arXiv2025

Global Counterfactual Directions

Despite increasing progress in development of methods for generating visual counterfactual explanations, especially with the recent rise of Denoising Diffusion Probabilistic Models, previous works consider them as an entirely local technique. In this work, we take the first step at globalizing them. Specifically, we discover that the latent space of Diffusion Autoencoders encodes the inference process of a given classifier in the form of global directions. We propose a novel proxy-based approach that discovers two types of these directions with the use of only single image in an entirely black-box manner. Precisely, g-directions allow for flipping the decision of a given classifier on an entire dataset of images, while h-directions further increase the diversity of explanations. We refer to them in general as Global Counterfactual Directions (GCDs). Moreover, we show that GCDs can be naturally combined with Latent Integrated Gradients resulting in a new black-box attribution method, while simultaneously enhancing the understanding of counterfactual explanations. We validate our approach on existing benchmarks and show that it generalizes to real-world use-cases.

cs.LG↗

arXiv2024

Interpretable Machine Learning for Survival Analysis

With the spread and rapid advancement of black box machine learning models, the field of interpretable machine learning (IML) or explainable artificial intelligence (XAI) has become increasingly important over the last decade. This is particularly relevant for survival analysis, where the adoption of IML techniques promotes transparency, accountability and fairness in sensitive areas, such as clinical decision making processes, the development of targeted therapies, interventions or in other medical or healthcare related contexts. More specifically, explainability can uncover a survival model's potential biases and limitations and provide more mathematically sound ways to understand how and which features are influential for prediction or constitute risk factors. However, the lack of readily available IML methods may have deterred medical practitioners and policy makers in public health from leveraging the full potential of machine learning for predicting time-to-event data. We present a comprehensive review of the limited existing amount of work on IML methods for survival analysis within the context of the general IML taxonomy. In addition, we formally detail how commonly used IML methods, such as such as individual conditional expectation (ICE), partial dependence plots (PDP), accumulated local effects (ALE), different feature importance measures or Friedman's H-interaction statistics can be adapted to survival outcomes. An application of several IML methods to real data on data on under-5 year mortality of Ghanaian children from the Demographic and Health Surveys (DHS) Program serves as a tutorial or guide for researchers, on how to utilize the techniques in practice to facilitate understanding of model decisions or predictions.

stat.ML↗

arXiv2024

The Dark Patterns of Personalized Persuasion in Large Language Models: Exposing Persuasive Linguistic Features for Big Five Personality Traits in LLMs Responses

This study explores how the Large Language Models (LLMs) adjust linguistic features to create personalized persuasive outputs. While research showed that LLMs personalize outputs, a gap remains in understanding the linguistic features of their persuasive capabilities. We identified 13 linguistic features crucial for influencing personalities across different levels of the Big Five model of personality. We analyzed how prompts with personality trait information influenced the output of 19 LLMs across five model families. The findings show that models use more anxiety-related words for neuroticism, increase achievement-related words for conscientiousness, and employ fewer cognitive processes words for openness to experience. Some model families excel at adapting language for openness to experience, others for conscientiousness, while only one model adapts language for neuroticism. Our findings show how LLMs tailor responses based on personality cues in prompts, indicating their potential to create persuasive content affecting the mind and well-being of the recipients.

cs.CL↗

arXiv2024

Resistance Against Manipulative AI: key factors and possible actions

If AI is the new electricity, what should we do to keep ourselves from getting electrocuted? In this work, we explore factors related to the potential of large language models (LLMs) to manipulate human decisions. We describe the results of two experiments designed to determine what characteristics of humans are associated with their susceptibility to LLM manipulation, and what characteristics of LLMs are associated with their manipulativeness potential. We explore human factors by conducting user studies in which participants answer general knowledge questions using LLM-generated hints, whereas LLM factors by provoking language models to create manipulative statements. Then, we analyze their obedience, the persuasion strategies used, and the choice of vocabulary. Based on these experiments, we discuss two actions that can protect us from LLM manipulation. In the long term, we put AI literacy at the forefront, arguing that educating society would minimize the risk of manipulation and its consequences. We also propose an ad hoc solution, a classifier that detects manipulation of LLMs - a Manipulation Fuse.

cs.HC↗

arXiv2024

Big Tech influence over AI research revisited: memetic analysis of attribution of ideas to affiliation

There exists a growing discourse around the domination of Big Tech on the landscape of artificial intelligence (AI) research, yet our comprehension of this phenomenon remains cursory. This paper aims to broaden and deepen our understanding of Big Tech's reach and power within AI research. It highlights the dominance not merely in terms of sheer publication volume but rather in the propagation of new ideas or memes. Current studies often oversimplify the concept of influence to the share of affiliations in academic papers, typically sourced from limited databases such as arXiv or specific academic conferences. The main goal of this paper is to unravel the specific nuances of such influence, determining which AI ideas are predominantly driven by Big Tech entities. By employing network and memetic analysis on AI-oriented paper abstracts and their citation network, we are able to grasp a deeper insight into this phenomenon. By utilizing two databases: OpenAlex and S2ORC, we are able to perform such analysis on a much bigger scale than previous attempts. Our findings suggest that while Big Tech-affiliated papers are disproportionately more cited in some areas, the most cited papers are those affiliated with both Big Tech and Academia. Focusing on the most contagious memes, their attribution to specific affiliation groups (Big Tech, Academia, mixed affiliation) seems equally distributed between those three groups. This suggests that the notion of Big Tech domination over AI research is oversimplified in the discourse.

physics.soc-ph↗