On 16 August 2026, Dr. Tome Eftimov, coordinator of the AutoLearn-SI project, delivered an invited talk at the AutoAI Meets Foundation Models (AutoAI-FM) Workshop, co-located with IJCAI-ECAI 2026 in Bremen, Germany.

His talk, “Trust, Afford, Use: An AutoAI Lens on Foundation-Model Evaluation and Selection,” explored how established AutoAI methods can support the reliable and efficient evaluation and selection of foundation models. The presentation addressed three central challenges: ensuring trustworthy model rankings, reducing evaluation costs and benchmark redundancy, and selecting the most suitable model for each task or instance.

The talk brought together research on multilingual text-embedding evaluation, tabular foundation models, prompt selection through combinatorial optimization, LLM routing, and the analysis of model design choices.

Special recognition goes to Ana Gjorgjevikj, Denica Kjorvezir, Gjorgjina Cenikj, and Ana Nikolikj, as well as all other researchers involved in the presented studies.