News
Two papers accepted at NeurIPS 2026. Routing Subspaces: Auditing Evaluation-to-Deployment Mismatch in Fine-Tuned Language Models is in the main track. “Optimal transport methods for computing similarity between plankton communities and forecasting with GNNs”, on marine monitoring data, is at the Geometric Distributional Deep Learning workshop.
Challenges in Deep Learning-Based Small Organ Segmentation: A Benchmarking Perspective for Medical Research with Limited Datasets published in Biomedical Signal Processing and Control.
The group hosted a workshop on AI for marine monitoring and sensing technologies.
A paper on signal processing of multimodal data accepted at the NeuroMM workshop at ACM Multimedia 2026.
Rebecca De Rosa joined the group as a Research Assistant.
Three papers accepted at ICML 2026 workshops. At the Mechanistic Interpretability workshop, Decoded but Unused: Instruction Tuning Routes Moral Framing into the Judgment Readout and A Path Already Walked: On Inheriting Network-Neuroscience Tools for Mechanistic Interpretability. At Philosophy Meets Machine Learning, Self-Reports Do Not Identify Self-Models: An Identifiability Test for Counterfactual Reports, presented as an oral.
Faithful or Findable? Evaluating LLM-Generated Metadata for RDF Dataset Search accepted at the SynthIR workshop at SIGIR 2026.
Why Neighborhoods Matter: Traversal Context and Provenance in Agentic GraphRAG accepted at the joint GenAIK–NORA workshop at IJCAI 2026.
Luigi Colluto and Manish Raj Moriche joined the group as Undergraduate Researchers.
PIPER: Content-Based Table Search via Profiling and LLM-Generated Pseudoqueries accepted at DEXA 2026, the International Conference on Database and Expert Systems Applications.
Data-driven warranty modeling for multi-state deteriorating products with stochastic repair times published in Reliability Engineering & System Safety.
Ane Cathrine Holst Merrild joined the group as a PhD Fellow, and Vladimír Jakabovic and Bartosz Bobkiewicz as Undergraduate Researchers.
Three papers accepted at ICSA 2026 workshops: CAKE: Cloud Architecture Knowledge Evaluation of Large Language Models at KDA-AI, A Reference Architecture for Agentic Hybrid Retrieval in Dataset Search at SAML, and Architecture Without Architects: How AI Coding Agents Shape Software Architecture at SAGAI.
Tim Lukas Adam joined the group.
Riccardo Terrenzi joined the group as a PhD Fellow.
Beyond Major Floods: Deep Learning for Detecting Shallow Water Inundation in Agricultural Areas presented at KES 2025.
RECOVER was granted: a Danish–German project on the resilience of coastal ecosystems in the western Baltic Sea, using microalgae as bioindicators and AI on environmental data towards a digital twin of the south-western Baltic. Funded by the EU’s Interreg Deutschland–Danmark programme, October 2025 to September 2028.
A Big Data Analytics System for Predicting Suicidal Ideation in Real-Time Based on Social Media Streaming Data published at DATA 2025, the International Conference on Data Science, Technology and Applications.
Toygar Tanyel visited the group on a travel grant from the Danish Data Science Academy (DDSA).
POE-ML: An Automated Pipeline for Optimization and Evaluation of Machine Learning accepted at an ICSA 2025 workshop.
The group joined the DIREC project Multimodal Data Processing of Earth Observation Data as a partner, and leads its work at SDU. The project builds processing pipelines for Earth observation data with Danish public organisations, to strengthen the data they use for environmental decisions.
Phongsakon Mark Konrad joined the group.
Remaining useful lifetime prediction for predictive maintenance in manufacturing published in Computers & Industrial Engineering.
Serkan Ayvaz joined SDU, and the group started at the Centre for Industrial Software.