LFLucie Flek

Lucie Flek

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Lucie Flek has a total of 75 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Shapes are not enough: CONSERVAttack and its use for finding vulnerabilities and uncertainties in machine learning applications

    by Philip Bechtle, Lucie Flek, Philipp Alexander Jung, Akbar Karimi, Timo Saala, Alexander Schmidt, Matthias Schott, Philipp Soldin, Christopher Wiebusch, Ulrich Willemsen

    In High Energy Physics, as in many other fields of science, the application of machine learning techniques has been crucial in advancing our understanding of fundamental phenomena. Increasingly, deep learning models are applied to analyze both simulated and experimental data. In most experiments, a rigorous regime of testing for physically motivated systematic uncertainties is in place. The numerical evaluation of these tests for differences between the data on the one side and simulations on the other side quantifies the effect of potential sources of mismodelling on the machine learning output. In addition, thorough comparisons of marginal distributions and (linear) feature correlations between data and simulation in “control regions” are applied. However, the guidance by physical motivation, and the need to constrain comparisons to specific regions, does not guarantee that all possible sources of deviations have been accounted for. We therefore propose a new adversarial attack – the CONSERVAttack – designed to exploit the remaining space of hypothetical deviations between simulation and data after the above mentioned tests. The resulting adversarial perturbations are consistent within the uncertainty bounds — evading standard validation checks — while successfully fooling the underlying model. We further propose strategies to mitigate such vulnerabilities and argue that robustness to adversarial effects must be considered when interpreting results from deep learning in particle physics.

  • Zoom In Disparities in Healthcare LLM Q&A

    by Ipek Baris Schlicht, Burcu Sayin, Zhixue Zhao, Frederik M. Labonté, Cesare Barbera, Marco Viviani, Paolo Rosso, Lucie Flek

  • Open Access

    Structured Multi-extraction Prompting Improves LLM-Based Biomedical Relation Extraction Without Fine-Tuning

    by Frederik Labonté, Suraj Giri, Claudius Herbrik, Lucie Flek

    Annotated text data is essential for structuring biomedical knowledge. State-of-the-art machine learning models achieve F1 scores >90% for named entity recognition (NER), but relation extraction (RE) remains challenging, with F1 scores of $$\tilde{6}$$ 6 ~ 0% (up to 75–80% on binary datasets using additional strategies). These models require fine-tuning on costly manually annotated datasets. This paper investigates structured multi-extraction prompting as a technique for enabling Large Language Models (LLMs) such as Mistral and OpenAI models to perform biomedical relation extraction (RE) without fine-tuning, reducing manual annotation burdens. We evaluate this approach across zero-shot and few-shot settings and compare extracting relations individually versus in batches, assessing whether multi-extraction with structured generation can serve as a viable, cost-effective alternative to fine-tuned models. Our results show that multi-extraction prompting with structured generation significantly boosts LLM performance on biomedical relation extraction (RE) tasks, roughly matching fine-tuned models on binary classification in zero-shot settings and approaching their performance on multi-class tasks with few-shot examples but the gap remains larger in those cases. By providing the model with a list of entities and leveraging structured, dynamic generation, we harness LLMs’ in-context learning to enhance both efficiency and accuracy. This method improves performance by up to 19% while reducing token usage by up to 550%, lowering computational costs comparing it to a single extraction baseline. In summary, structured multi-extraction prompting enables modern LLMs to perform binary RE tasks without fine-tuning, matching or approaching the performance of small specialised models, while substantially reducing token usage and computational cost—making them viable annotators for low-resource scenarios where little to no training data exists.

  • Open Access

    MiniFool: Physics-Constraint-Aware Minimizer-Based Adversarial Attacks in Deep Neural Networks

    by Lucie Flek, Oliver Janik, Philipp Alexander Jung, Akbar Karimi, Timo Saala, Alexander Schmidt, Matthias Schott, Philipp Soldin, Matthias Thiesmeyer, Christopher Wiebusch, et al.

    In this paper, we present a new algorithm, , that implements physics-inspired adversarial attacks for testing neural network-based classification tasks in particle and astroparticle physics. While we initially developed the algorithm for the search for astrophysical tau neutrinos with the IceCube Neutrino Observatory, we apply it to further data from other science domains, thus demonstrating its general applicability. Here, we apply the algorithm to the well-known MNIST data set and furthermore, to public data from the CMS Experiment at the Large Hadron Collider. The algorithm is based on minimizing a cost function that combines a $$\chi ^2$$ χ 2 based test-statistic with the deviation from the desired target score. The test statistic quantifies the probability of the perturbations applied to the data based on the experimental uncertainties. For our studied use cases, we find that the likelihood of a flipped classification differs for both the initially correctly and incorrectly classified events. When testing changes of the classifications as a function of an attack parameter that scales the experimental uncertainties, the robustness of the network decision can be assessed. Furthermore, this allows testing the robustness of the classification of unlabeled experimental data.


  • Open Access

    (Re-)Thinking Empathy's Materiality in HCI

    by Sophia Ppali, Mireia Yurrita, Alice Vitali, Alok Debnath, Lucie Flek, Andrea Cuadra, Sven Mayer, Michal Lahav, Tiffanie Horne, Aneesha Singh, et al.

  • Open Access

    CHARISMA: Character-Based Interaction Simulation with Multi-LLM Agents Toward Computational Social Psychology

    by Vahid Sadiri Javadi, Fryderyk Róg, Aksa Aksa, Johanne Trippas, Svitlana Vakulenko, Lucie Flek

    How people seek, request, and exchange information in social interactions is shaped by personality and situational context, connecting the fields of interactive information science and attribution theory in social psychology. In everyday life, people seek information to achieve goals, collaborate, and manage social conflicts. Understanding how individual traits and contextual factors influence information-seeking behavior remains a challenge. Recent advances with large language models (LLMs) enable the simulation of socially grounded information-seeking behaviors in realistic and controllable ways. We introduce CHARISMA, a simulation framework that uses LLMs to examine how personality traits and situational factors influence information seeking as a form of social behavior. CHARISMA leverages movie characters and public figures as personality anchors, drawing on LLMs’ knowledge to simulate human-like interaction. CHARISMA’s utility is demonstrated in two studies: (1) agreeable pairs resolve conflicts more successfully, and (2) low-agreeable agents compete for information, while high-agreeable agents cooperate through prosocial exchange.

  • Open Access

    MiniFool -- Physics-Constraint-Aware Minimizer-Based Adversarial Attacks in Deep Neural Networks

    by Lucie Flek, Oliver Janik, Philipp Alexander Jung, Akbar Karimi, Timo Saala, Alexander Schmidt, Matthias Schott, Philipp Soldin, Matthias Thiesmeyer, Christopher Wiebusch, et al.

    In this paper, we present a new algorithm, MiniFool, that implements physics-inspired adversarial attacks for testing neural network-based classification tasks in particle and astroparticle physics. While we initially developed the algorithm for the search for astrophysical tau neutrinos with the IceCube Neutrino Observatory, we apply it to further data from other science domains, thus demonstrating its general applicability. Here, we apply the algorithm to the well-known MNIST data set and furthermore, to Open Data data from the CMS experiment at the Large Hadron Collider. The algorithm is based on minimizing a cost function that combines a $χ^2$ based test-statistic with the deviation from the desired target score. The test statistic quantifies the probability of the perturbations applied to the data based on the experimental uncertainties. For our studied use cases, we find that the likelihood of a flipped classification differs for both the initially correctly and incorrectly classified events. When testing changes of the classifications as a function of an attack parameter that scales the experimental uncertainties, the robustness of the network decision can be quantified. Furthermore, this allows testing the robustness of the classification of unlabeled experimental data.

  • Open Access

    IKnow: Instruction-Knowledge-Aware Continual Pretraining for Effective Domain Adaptation

    by Tianyi Zhang, Florian Mai, Lucie Flek

    Continual pretraining promises to adapt large language models (LLMs) to new domains using only unlabeled test-time data, but naively applying standard self-supervised objectives to instruction-tuned models is known to degrade their instruction-following capability and semantic representations. Existing fixes assume access to the original base model or rely on knowledge from an external domain-specific database - both of which pose a realistic barrier in settings where the base model weights are withheld for safety reasons or reliable external corpora are unavailable. In this work, we propose Instruction-Knowledge-Aware Continual Adaptation (IKnow), a simple and general framework that formulates novel self-supervised objectives in the instruction-response dialogue format. Rather than depend- ing on external resources, IKnow leverages domain knowledge embedded within the text itself and learns to encode it at a deeper semantic level.

  • Do LLMs Provide Consistent Answers to Health-Related Questions Across Languages?

    by Ipek Baris Schlicht, Zhixue Zhao, Burcu Sayin, Lucie Flek, Paolo Rosso