Tome Eftimov, coordinator of the AutoLearn-SI project, has been recognized as an ICML 2026 Gold Reviewer, placing him among the top reviewers at one of the leading international AI conferences. This recognition highlights the importance of scientific community service and the role of high-quality peer review in advancing trustworthy and rigorous machine learning research.

Researchers from the Jožef Stefan Institute presented the FoodBench-QA shared task at the CL4Health Workshop collocated with LREC-COLING 2026, showcasing advances in trustworthy and semantically grounded AI for food and nutrition question answering.

Prof. Sašo Džeroski represented the ERA Chair AutoLearn-SI project at the 6th Presidential Forum on Artificial Intelligence, where the importance and impact of trustworthy AI initiatives and the Slovenian AI Factory (SLAIF) were highlighted at the national level.

An invited talk by Fernando Spadea explored AI-driven approaches for proactive risk prevention and systemic stability in decentralized finance ecosystems.

The Brave Conversations event, organized within the ERA Chair AutoLearn-SI activities, brought together researchers, students, and professionals to discuss whether humans shape technology or technology shapes society in the age of AI.

The AI and Data Privacy and Security Training at the Jožef Stefan Institute brought together 50+ participants for a hands-on exploration of decentralized, privacy-preserving AI and a policy discussion on the EU AI Act led by Oshani Seneviratne, Fernando Spadea, and Polona Pičman Štefančič.

We hosted an invited talk by Oshani Seneviratne (Rensselaer Polytechnic Institute**), presenting a forward-looking vision of decentralized AI ecosystems that are resilient, accountable, and user-centric.

Lars Kotthoff (University of St Andrews) delivered a guest lecture at the Jožef Stefan Institute on applying AI and Bayesian optimization to improve laser-induced graphene production, achieving up to twofold performance gains over existing methods. The talk highlighted the potential of automated machine learning in materials science and sparked engaging discussions with participants.

At Dnevi Slovenske Informatike (DSI) 2026, Tome Eftimov, together with Ana Nikolić and Matevž Ogrinc, presented insights from the AutoLearn-SI ERA Chair Project, showing that while AutoML is widely known, only 27% of stakeholders have experimented with it in practice.

They also shared findings from the DATA-TRUST project, revealing that only 27.3% routinely evaluate bias in AI workflows, and introduced the upcoming AutoML Conference 2026 in Ljubljana.

We are pleased to welcome Gianpaolo Torre, who joins our research group as an MSCA SQUASH Postdoctoral Researcher working on “Tensor Networks for Hybrid Classical–Quantum Technologies.”

His project explores tensor networks as efficient and interpretable models for sustainable AI and quantum information science, advancing scalable methods for hybrid classical–quantum computation.

Submissions are invited to the 1st International Workshop on Learning-assisted Algorithm Design (LEAD 2026) at IJCAI-ECAI 2026 (Bremen, Germany, Aug 15–17, 2026), focusing on learning-driven approaches for algorithm design, optimization, and automated decision systems.

AutoLearn-SI organized a hands-on workshop on LLMs for Administration at the Jožef Stefan Institute, bringing together administrative colleagues to explore how AI can support everyday work processes. Participants discussed real challenges and practical use cases, gaining insights into how LLMs can help automate parts of their daily workflows.

A one-day training event on May 5, 2026 at the Jožef Stefan Institute in Ljubljana will bring together students, researchers, and entrepreneurs to explore trustworthy AI, data privacy, and security through a scientific tutorial and policy discussion. The program includes sessions on knowledge-graph–powered decentralized personalization, federated and blockchain-based AI architectures, KG-grounded RAG pipelines, hands-on exercises, and a policy overview of the EU AI Act.

Our methodological advances are contributing to real-world impact by supporting collaboration with COST Action PerMediK on Big Data–driven personalized medicine in chronic kidney disease.

FoodBench-QA, accepted at LREC 2026, introduces a benchmark for grounded food and nutrition question answering, highlighting strong performance across tasks while emphasizing the need for trustworthy, explainable, and regulation-aware AI systems.

Two papers by AutoLearn-SI PhD students have been accepted at IEEE WCCI 2026 (IEEE CEC), highlighting cutting-edge research in optimization and graph benchmarking. The works reveal fundamental limits of structure-based representations, showing that similar problem structures do not always imply similar algorithm performance.

The AutoLearn-SI ERA Chair project at the Jožef Stefan Institute is hosting Dr. Marko Djukanović, a Marie Skłodowska-Curie postdoctoral fellow from the SMASH project, for a three-month research secondment. During his visit, he will collaborate with the team on methods for bias-aware selection of optimization problem instances using dimensionality reduction, unsupervised learning, and combinatorial optimization to support meta-learning approaches for explaining algorithm performance.

The coordinator of the AutoLearn-SI ERA Chair project, Tome Eftimov, delivered an invited talk at NEC Laboratories Europe titled “Leveraging Benchmarking Data for Automated Optimization/Machine Learning.” The talk presented how benchmarking data can support automated and interpretable algorithm design through representative benchmark instance selection and the concept of algorithmic footprints, which capture how algorithms interact with problem landscapes.

The Jožef Stefan International Postgraduate School (MPS), supported by the ERA Chair AutoLearn-SI, will host a Brave Conversations event on May 6, 2026 (09:00–12:00) at the Jožef Stefan Institute in Ljubljana, organized together with Dr. Oshani Seneviratne from Rensselaer Polytechnic Institute. The event is part of a global initiative that creates interactive spaces for open dialogue on emerging technologies and their societal impact, bringing together researchers, students, and professionals for small-group discussions on the future of AI and digital technologies.

The AutoLearn-SI group at the Jožef Stefan Institute hosted a research mobility visit within the SPECTRA Project, strengthening collaboration on AI-assisted analysis of environmental and food data while promoting FAIR data and open science practices.

As part of AutoLearn-SI, our project coordinator delivered a workshop on LLM-supported automated data analysis at JSI, highlighting ML pipeline fundamentals, the benefits and limitations of LLM-generated code, and practical examples of responsible AI use in research.

The AutoLearn-SI Science Communication Workshop brought together researchers from diverse fields to strengthen their skills in effective research dissemination, proposal writing, and responsible scientific communication in line with Horizon Europe priorities.

The AutoLearn-SI project announces that project coordinator Tome Eftimov is part of the organizing team of the GECCO 2026 Workshop on Large Language Models for and with Evolutionary Computation, exploring the synergy between large language models and evolutionary methods. The workshop welcomes extended abstracts and full papers addressing new approaches and applications at the intersection of AI and evolutionary computation.

Ana Nikolikj has been invited as the 18th lecturer in the AToNIIC Lecture Series, where she will present her work on modular optimization frameworks and benchmarking of black-box optimization algorithms. Her research advances more trustworthy and explainable evaluation methods in AutoML and optimization, contributing to AutoLearn-SI’s mission of developing reliable AI systems.

Our newly accepted paper introduces FoodyLLM, a domain-specialized large language model for nutrient estimation, traffic-light multi-label classification, and ontology-based food entity linking, significantly outperforming general-purpose LLMs across all tasks. The work highlights the importance of trustworthy benchmarking (AutoLearn-SI) and sustainability-aware model selection (AutoLLMSelect), demonstrating that smaller domain-adapted models can achieve higher accuracy with lower computational cost.

On the International Day of Women and Girls in Science, we proudly celebrate the outstanding women of the AutoLearn-SI ERA Chair whose leadership, research excellence, and innovation are advancing trustworthy AI, AutoML, NLP, and interdisciplinary science. Their dedication, talent, and collaboration embody our belief that diversity in science drives stronger ideas, impactful solutions, and a more responsible AI future. 💙

Join us on February 26, 2026 (10:00–12:00) at the Jožef Stefan Institute (JSI) for an interactive Science Communication workshop focused on effectively communicating research to the public, stakeholders, and the scientific community. The workshop is open to national and visiting international MSc and PhD students.

Ana Nikolikj will speak at DATA_FAIR 2026 in Ljubljana on practical strategies for selecting large language models while balancing accuracy, scale, and sustainability. Her talk highlights how academic AI research can be translated into real-world, responsible, and sustainability-aware machine learning practice.

AutoML has been officially recognized as a core AI component at the kick-off meeting of the Slovenian AI Factory, underscoring its strategic role in Slovenia’s national AI infrastructure. This recognition highlights the growing national visibility and impact of the AutoLearn-SI ERA Chair, positioning it as a key contributor to the Slovenian AI ecosystem.

Gjorgjina Cenikj, a member of the AutoLearn-SI research team at the Jožef Stefan Institute, will present an invited talk at the COST ROAR-NET WG5 / Task Force 2 meeting on Landscape Features and the Limits of Algorithm Selection. The talk, based on recent work published in Swarm and Evolutionary Computation, examines the generalization limits of landscape features in data-driven algorithm selection for continuous optimization.

This survey reviews recent advances in feature design for representing optimization problems, algorithms, and their interactions in single-objective continuous black-box optimization, highlighting their role in machine learning tasks such as algorithm selection, configuration, and problem classification. It also discusses current limitations of the state of the art and outlines promising directions for future research.

The AutoML Conference 2026 will be hosted in Ljubljana, Slovenia, the Green Capital of Europe, bringing the global AutoML community to a sustainable and vibrant research environment. The AutoLearn-SI project coordinator, Tome Eftimov, will serve as General Chair of AutoML 2026, reinforcing the project’s strong engagement with the international AutoML ecosystem.

Two new national projects, DATA-TRUST and AI4Food, have been funded by the Slovenian Research and Innovation Agency (ARIS), marking an important milestone for the development of trustworthy AI and AI for food science in Slovenia. This success significantly strengthens the long-term sustainability of the AutoLearn-SI ERA Chair by ensuring scientific continuity, strategic alignment, and stable national support as we move into 2026.

While our ERA Chair holder is representing AutoLearn-SI abroad, the rest of the team sends warm holiday wishes and looks forward to an inspiring and productive year ahead. 🎄✨

Our tutorial on recent advances in meta-features for representing and benchmarking black-box single-objective continuous optimization has been accepted for presentation at GECCO 2026.

Our ERA Chair holder presented recent AutoLearn-SI research on sustainable benchmarking as an invited speaker at the International Conference on Machine and Computing Technologies for Sustainable Development (Manila, December 11–13, 2025).

FoodBench-QA 2026 is a shared task on grounded food and nutrition question answering, challenging systems to perform nutrient estimation, FSA traffic-light prediction, and food entity recognition/linking using structured dietary resources. Hosted at CL4Health @ LREC 2026, the competition provides realistic food-related queries and invites participants to submit system papers following the evaluation phase.

We invite researchers to submit their work to the 5th AutoCIS Special Session at WCCI 2026, focusing on advancing automation in computational intelligence through AutoML, evolutionary methods, and real-world optimisation applications.

Our ERA Chair Holder delivered a keynote on Automated Optimization and Automated Machine Learning at the ICIDA 2025 conference in Kolkata, India, speaking alongside experts in data analytics.

Ana Nikolikj presented her work “Benchmarking Algorithm Footprints: Explainable Insights into Algorithm Success and Failure” at the Science of Benchmarking and Evaluating AI Workshop at EurIPS 2025 in Copenhagen. Her talk highlighted how algorithm footprints reveal why algorithms succeed or fail across different search-space regions, advancing transparency in optimization and AutoML.

We are proud to announce that Dr. Gjorgjina Cenikj has successfully defended her PhD, marking an important milestone for her career and a great achievement for our project team.

The AutoLearn-SI Exploratory Workshop brought together 75 participants over two days, creating an inspiring environment for valuable exchanges and insightful discussions.

Day 2 of the AutoLearn-SI workshop focused on future-oriented perspectives in optimization and AI, featuring insightful talks by Dr. Carola Doerr on connecting practical needs with fundamental research in black-box optimization and Associate Prof. Niki van Stein on leveraging large language models for automatic algorithm design.

The first day of the AutoLearn-SI workshop at JSI was a great success, featuring excellent talks by Dr. Ana Kostovska, Dr. Jan van Rijn, and Dr. Sebastian Rojas Gonzalez, and fostering active discussions on benchmarking, AutoML, and Bayesian optimization.

Our project coordinator, Tome Eftimov, will give the next talk in the AutoML Seminar on November 20th at 3:00 PM CET, titled “Benchmarking Beyond Statistics: Data-Driven Footprints for Explainable Black-Box Optimization.” The AutoML Seminar is a free, open platform that hosts leading researchers to discuss the latest advances in Automated Machine Learning.

The AutoLearn-SI Two-Day Exploratory Workshop will bring together leading researchers and innovators to discuss the next generation of Automated Machine Learning (AutoML) and Automated Optimization (AutoOpt) — from benchmarking and learning methodologies to large language model (LLM) integration in optimization.

Our project coordinator, Dr. Tome Eftimov, received the Information Society Award 2025 for the best research and development work in computer science in Slovenia, recognizing his leadership in advancing transparent, explainable, and sustainable AI through the AutoLearn-SI ERA Chair project.

The AutoLearn-SI ERA Chair project held its first Advisory Board meeting and Stakeholder Workshop, presenting early results and revealing that only 27% of participants currently use AutoML—highlighting the project’s timely role in advancing automated AI adoption across Slovenia.

Our ERA Chair holder presented the AutoLearn-SI vision and green computing advances in AutoML at the DATA_FAIR Meetup, connecting academia with industry to promote sustainable AI innovation.

ERA Chair Holder Eva Tuba Keynote at the Slovenian AI Conference 2025

Prof. Eva Tuba, ERA Chair holder of the Horizon Europe AutoLearn-SI project, delivered a keynote lecture “Automated Optimization and Machine Learning: Challenges and Opportunities in AI Research” at the Slovenian Conference on Artificial Intelligence 2025, held on 8–9 October at the Jožef Stefan Institute as part of the Information Society Conference.

Two ERA Chair members, Asst. prof. Dr. Tome Eftimov and Prof. Sašo Džeroski, have been recognized in the Elsevier Data Repository of the World’s Top 2% Cited Researchers 2024, highlighting their outstanding single-year and career-long scientific impact.

🔔 The AutoLearn-SI Stakeholder Workshop (Oct 10, 2025, Jožef Stefan Institute, Ljubljana) will bring together stakeholders from academia, industry, government, and NGOs to shape the future of AI applications through AutoML and AutoOpt.

Our team presented three contributions at the International Conference on AI for Science (Discovery Science 2025), highlighting advances in explainable benchmarking, graph-based performance prediction, and analysis of PSO-X components. We thank all collaborators and the community for the valuable discussions and feedback during the event.

The AutoLearn-SI team led a workshop at IMEKOFOODS 2025, showcasing how AutoML and LLMs can advance food data analysis and interoperability.

Prof. dr. Riste Stojanov, from our associated partner FCSE, presented the vision and research methods of AutoLearn-SI at the COST Action PerMediK Workshop & 7th Management Committee Meeting, highlighting the key role of AutoML in shaping the future of biomedical research.

Prof. Dr. Sašo Džeroski delivered the keynote “AI for Science” at ECML-PKDD 2025, highlighting advances in explainable machine learning, learning from limited data, automated scientific modeling, and semantic technologies, with the AutoLearn-SI project prominently featured.

Ana Nikolikj presented our work on explainable algorithm performance, showcasing how algorithm footprints and meta-feature interactions can be used to transparently benchmark multi-label classification methods at ECML-PKDD 2025 in Porto.

The AutoLearn-SI project was featured in the Slovenian AI Factory presentation at the Austrian-Slovenian HPC Meeting 2025, highlighting Slovenia’s AI and HPC innovations.

Our article, accepted by Knowledge-Based Systems, presents xLLMBench, a transparent framework for ranking LLMs using multi-criteria decision-making. Evaluated on HuggingFace and HELM Classic leaderboards, it enables application-specific benchmarking, revealing how rankings shift under different metrics for transparent and reproducible insights.

Prof. Sašo Džeroski, ERA Chair member of the AutoLearn-SI project, joined a national radio podcast on RTV SLO to discuss the role of AI in society and its environmental impact. The conversation explored AI’s accuracy, applications, future developments, and its potential to reduce—or worsen—our carbon footprint.

Jakub Kůdela from Brno University of Technology delivered an invited talk on bridging mathematical programming and evolutionary computation, highlighting benchmarking methods for optimization algorithms and real-world challenges relevant to AutoOpt and AutoML.

🚀 We’re proud to share that Prof. Dr. Sašo Džeroski, ERA Chair member of AutoLearn-SI, will deliver a keynote at ECML PKDD 2025 on Artificial Intelligence for Science. His talk will explore explainable AI, learning from limited data, and semantic technologies for open and reproducible science.

ChatGPT said: We were pleased to host Yihang Lu, an MSCA doctoral researcher from Sorbonne University, who gave an insightful talk on unsupervised learning for food traceability. The session sparked engaging discussion around future directions, particularly how multi-view learning could support AutoML techniques in food applications.

This study presents an in-depth analysis of module importance in modular CMA-ES using exploratory data analysis and large-scale benchmarking across the BBOB suite. We show that understanding and selectively tuning key module interactions—quantified via f-ANOVA—can rival or even surpass the performance of the best full-configured solver, especially in high-dimensional problems.

Our short paper "Quantifying Module Interactions in the PSO-X Framework", co-authored by several ERA Chair AutoLearn-SI members, has been accepted for presentation in the non-archival track at AutoML 2025.

The Jožef Stefan Institute is offering two full-time postdoctoral positions in Automated Machine Learning and Optimization (AutoML/AutoOPT), starting June 1, 2026, within the Horizon Europe-funded ERA Chair project AutoLearn-SI, with applications open until December 31, 2025.

We’re excited to host a ROAR-NET COST Action Code Fest hub in Ljubljana as part of a global 48-hour event focused on collaborative problem modelling and optimization.

Our paper "Explainable Insights on Algorithm Performance" by Ana Nikolikj, Mario Andrés Muñoz, and Tome Eftimov has been accepted to the ECML-PKDD 2025 Nectar Track. Originally published in Swarm and Evolutionary Computation, the work introduces a general methodology for explaining algorithm performance, initially applied to black-box optimization but now extended to multi-label classification in machine learning.

An open-source large language model for food named-entity linking, FoodSEM advances multi-label classification and sustainable AI, showcasing the synergy between the AutoLLMSelect and AutoLearn-SI projects.

Our team announced open PhD positions and presented new research on adaptive run estimation for sustainable benchmarking at GECCO 2025.

The third day of GECCO 2025 featured impactful contributions from the ERA Chair AutoLearn-SI team, showcasing advances in understanding algorithm behavior through landscape analysis, search trajectory clustering, and geometric learning with GNNs for performance prediction in black-box optimization.

The ERA Chair team delivered a tutorial at GECCO 2025 on advanced feature representations in black-box optimization.

The establishment of AutoLearn-SI has been highlighted in SIGEVOlution, the newsletter of the ACM Special Interest Group on Genetic and Evolutionary Computation (SIGEVO). This recognition underlines AutoLearn-SI’s emerging role in advancing automated learning and optimization research.

Slovenia’s national AI reforms are built on secure infrastructure, strategic frameworks, and hands-on pilot projects under the Slovenian AI Factory.

At the close of IEEE CEC 2025, our team presented three research papers and chaired the special session “Automating Computational Intelligence Systems: Trends, Challenges, and Real-World Applications.”

The Jožef Stefan Institute’s Horizon Europe AutoLearn-SI project is seeking motivated doctoral candidates to conduct cutting-edge research in automated machine learning and optimisation as part of its new ERA Chair group.

This week in Hangzhou, the ERA Chair team will unveil new AutoLearn-SI research at the IEEE Congress on Evolutionary Computation.

Applications are now open for fully funded, three-month SPECIES internships at our lab—complete with additional support from us—so submit your proposal and choose us as your host!

The AutoLearn-SI ERA Chair—created to drive cutting-edge automated machine-learning and optimization research in Slovenia—has been spotlighted in the latest SLAIS newsletter, underscoring its national and European significance.

Dr Ana Gjorgjevikj has won a Marie Skłodowska-Curie Postdoctoral Fellowship for her AutoLLMSelect project, which—under the guidance of Asst. Prof. Tome Eftimov and Prof. Barbara Koroušić Seljak—will extend our ERA Chair’s automated-learning research into large language models, securing vital new funding and impact; follow AutoLLMSelect for future results.

ERA Chair members Sašo Džeroski and Tome Eftimov showcased advances in AI-driven discovery and benchmark-guided experimental design at the Ruđer Bošković Institute’s AI for Science workshop in Zagreb.

Our project coordinator presented cutting-edge methods for leveraging benchmarking data to automate model optimization at the AI Workshop held by the Faculty of Natural Sciences and Mathematics, University of Banja Luka. Attendees engaged in a dynamic discussion on standardizing benchmarks, automating hyperparameter tuning, and fostering future academic–industry collaborations.

Ana Kostovska presented at the Slovenian Data Science community’s DATA FAIR event in Ljubljana, uniting industry and academia. She demonstrated how graph structures can predict the best-performing algorithm for any learning task and sparked an engaging discussion with the audience.

At the CASSINI Hackathon’s Big Ideas Campaign, our Project Coordinator showcased how AI is transforming the food and health sectors and will next week mentor teams to refine their projects and spark innovative solutions, with our thanks to the organizers.

Ana Nikolikj demonstrated at COSEAL 2025 that in modular CMA-ES, module importance is driven primarily by multi-modality in low dimensions and by a blend of scalability, multi-modality, and global structure in high dimensions, underscoring the need for problem-specific module configurations.

At COSEAL 2025 in Porto, Gjorgjina Cenikj presented her study probing the adequacy of current landscape-characterization techniques for algorithm selection in single-objective continuous optimization.

We’re proud to announce the publication of “A Learning Search Algorithm for the Restricted Longest Common Subsequence Problem” in Expert Systems with Applications, where we introduce probabilistic partial‐solution evaluation and neural‐guided beam search heuristics to boost RLCS performance. Our empirical explainability analysis uncovers the key feature combinations driving algorithmic success and showcases the broad applicability of AutoLearn-SI methods across combinatorial optimization challenges.

The CIIT2025 session in Strumica on April 25, 2025, featured two presentations on benchmarking-driven optimization and user-centric LLM ranking, followed by a panel exploring human-centric automation, transparency, and interoperability.

At Dnevi slovenske informatike 2025, AutoLearn-SI was featured in both the final panel and project exhibition, highlighting key challenges in responsible AI, including transparency, ethics, and human oversight.

At the loveHR 2025 Summit, our project coordinator highlighted how automated learning techniques can transform workforce development across recruitment, talent management, and organizational strategy.

Asst. Prof. Dr. Eva Tuba, ERA Chair Holder of AutoLearn-SI, delivered a keynote on AI-driven automated optimization at the 6th Doctoral Symposium on Computational Intelligence, highlighting cutting-edge research in single-objective optimization.

At JSI Open Days 2025, Asst. Prof. Dr. Tome Eftimov presented the AutoLearn-SI project, highlighting its data-driven research and global collaboration opportunities in AI and optimization.

We’re excited to present two regular papers and two posters at GECCO 2025, showcasing our latest research in stochastic optimization, multi-objective performance modeling, algorithm behavior analysis, and GNN-based performance prediction.

Three of our papers - two in the regular track and one in the journal-to-conference track - have been accepted for presentation at IEEE CEC 2025, showcasing our latest advances in algorithm analysis, visualization, and benchmarking.

Congratulations to Dr. Ana Kostovska on her successful PhD defense, with research that lays a key foundation for the AutoLearn-SI project in benchmarking and optimization.

From Paris to JSI: Breakthroughs in Low-Discrepancy Point Sets with Dr. Carola Doerr

The official kick-off meeting for AutoLearn-SI, a HorizonEU-funded project at Jozef Stefan Institute (JSI), Slovenia. This project establishes an ERA Chair research group to advance AutoML and AutoOPT, integrating cutting-edge automation into research and education.