I am involved in teaching the following courses:

• Machine Learning (ML)
  Lecturer & Course Assistant; since spring 2024
• Introduction to Programming (INTROPROG / MAR-IP)
  Lecturer & Course Assistant; since fall 2024
• Project management and tools for health informatics (PROHI)
  Lecturer & Course Assistant; since fall 2024

Past courses I have been involved in:

• Principles and Foundations of Artificial Intelligence (PFAI)
  Administration, Lecturer & Course Assistant; Fall 2023

A selection from Stockholm University publication database

• Mind the gap: from plausible to valid self-explanations in large language models

  2025. Korbinian Robert Randl (et al.). Machine Learning 114 (10)

  Article

  This paper investigates the reliability of explanations generated by large language models (LLMs) when prompted to explain their previous output. We evaluate two kinds of such self-explanations (SE)—extractive and counterfactual—using state-of-the-art LLMs (1B to 70B parameters) on three different classification tasks (both objective and subjective). In line with Agarwal et al. (Faithfulness versus plausibility: On the (Un)reliability of explanations from large language models. 2024. https://doi.org/10.48550/arXiv.2402.04614), our findings indicate a gap between perceived and actual model reasoning: while SE largely correlate with human judgment (i.e. are plausible), they do not fully and accurately follow the model’s decision process (i.e. are not faithful). Additionally, we show that counterfactual SE are not even necessarily valid in the sense of actually changing the LLM’s prediction. Our results suggest that extractive SE provide the LLM’s “guess” at an explanation based on training data. Conversely, counterfactual SE can help understand the LLM’s reasoning: We show that the issue of validity can be resolved by sampling counterfactual candidates at high temperature—followed by a validity check—and introducing a formula to estimate the number of tries needed to generate valid explanations. This simple method produces plausible and valid explanations that offer a 16 times faster alternative to SHAP on average in our experiments.

  Read more about Mind the gap

• Evaluating the Reliability of Self-Explanations in Large Language Models

  2025. Korbinian Robert Randl (et al.). Discovery Science, 36-51

  Conference

  This paper investigates the reliability of explanations generated by large language models~(LLMs) when prompted to explain their previous output. We evaluate two kinds of such self-explanations -- extractive and counterfactual -- using three state-of-the-art LLMs (2B to 8B parameters) on two different classification tasks (objective and subjective).

  Our findings reveal, that, while these self-explanations can correlate with human judgement, they do not fully and accurately follow the model's decision process, indicating a gap between perceived and actual model reasoning.

  We show that this gap can be bridged because prompting LLMs for counterfactual explanations can produce faithful, informative, and easy-to-verify results. These counterfactuals offer a promising alternative to traditional explainability methods (e.g. SHAP, LIME), provided that prompts are tailored to specific tasks and checked for validity.

  Read more about Evaluating the Reliability of Self-Explanations in Large Language Models

• CICLe: Conformal In-Context Learning for Largescale Multi-Class Food Risk Classification

  2024. Korbinian Robert Randl (et al.). Findings of the Association for Computational Linguistics, 7695-7715

  Conference

  Contaminated or adulterated food poses a substantial risk to human health. Given sets of labeled web texts for training, Machine Learning and Natural Language Processing can be applied to automatically detect such risks. We publish a dataset of 7,546 short texts describing public food recall announcements. Each text is manually labeled, on two granularity levels (coarse and fine), for food products and hazards that the recall corresponds to. We describe the dataset and benchmark naive, traditional, and Transformer models. Based on our analysis, Logistic Regression based on a tf-idf representation outperforms RoBERTa and XLM-R on classes with low support. Finally, we discuss different prompting strategies and present an LLM-in-the-loop framework, based on Conformal Prediction, which boosts the performance of the base classifier while reducing energy consumption compared to normal prompting.

  Read more about CICLe: Conformal In-Context Learning for Largescale Multi-Class Food Risk Classification

• Early prediction of the risk of ICU mortality with Deep Federated Learning

  2023. Korbinian Robert Randl (et al.). 2023 IEEE 36th International Symposium on Computer-Based Medical Systems (CBMS), 706-711

  Conference

  Intensive Care Units usually carry patients with a serious risk of mortality. Recent research has shown the ability of Machine Learning to indicate the patients’ mortality risk and point physicians toward individuals with a heightened need for care. Nevertheless, healthcare data is often subject to privacy regulations and can therefore not be easily shared in order to build Centralized Machine Learning models that use the combined data of multiple hospitals. Federated Learning is a Machine Learning framework designed for data privacy that can be used to circumvent this problem. In this study, we evaluate the ability of deep Federated Learning to predict the risk of Intensive Care Unit mortality at an early stage. We compare the predictive performance of Federated, Centralized, and Local Machine Learning in terms of AUPRC, F1-score, and AUROC. Our results show that Federated Learning performs equally well as the centralized approach (for 2, 4, and 8 clients) and is substantially better than the local approach, thus providing a viable solution for early Intensive Care Unit mortality prediction. In addition, we demonstrate that the prediction performance is higher when the patient history window is closer to discharge or death. Finally, we show that using the F1-score as an early stopping metric can stabilize and increase the performance of our approach for the task at hand.

  Read more about Early prediction of the risk of ICU mortality with Deep Federated Learning

• A Workflow for Creating Multimodal Machine Learning Models for Metastasis Predictions in Melanoma Patients

  2025. Franco Rugolon (et al.). Machine Learning and Principles and Practice of Knowledge Discovery in Databases, 87-102

  Conference

  Melanoma is the most common form of skin cancer, responsible for thousands of deaths annually. Novel therapies have been developed, but metastases are still a common problem, increasing the mortality rate and decreasing the quality of life of those who experience them. As traditional machine learning models for metastasis prediction have been limited to the use of a single modality, in this study we aim to explore and compare different unimodal and multimodal machine learning models to predict the onset of metastasis in melanoma patients to help clinicians focus their attention on patients at a higher risk of developing metastasis, increasing the likelihood of an earlier diagnosis.

  We use a patient cohort derived from an Electronic Health Record, and we consider various modalities of data, including static, time series, and clinical text. We formulate the problem and propose a multimodal ML workflow for predicting the onset of metastasis in melanoma patients. We evaluate the performance of the workflow based on various classification metrics and statistical significance. The experimental findings suggest that multimodal models outperform the unimodal ones, demonstrating the potential of multimodal ML to predict the onset of metastasis.

  Read more about A Workflow for Creating Multimodal Machine Learning Models for Metastasis Predictions in Melanoma Patients

Show all publications by Korbinian Randl at Stockholm University