Zuse School ELIZA at IJCAI 2026: Learning, Networking, and Presenting Research in Bremen

A strong start: agentic AI and a Diversity & Inclusion lunch

The conference opened with a keynote from Prof. Nick Jennings CB FRS FREng, who spoke on the past, present, and future of agentic AI. As ELIZA student Stefani Kecman recounted, the talk argued that agentic AI is not a new idea, but that today’s efforts are focused on making human-agent collaboration more flexible and effective, improving communication between independently developed agents, and increasing AI accountability. Jennings offered a memorable framing for the field’s next step: “A paradigm shift will not come from building more smart agents, but intelligent societies.”

The same day brought one of the conference’s highlights off the technical track: a Diversity and Inclusion Lunch held at the historic Bremen Town Hall. ELIZA students used the occasion to network with the wider AI community and discuss the impact of AI on education, human creativity, and society.

Zuse Schools come together, with DAAD

IJCAI also gave the Zuse Schools a chance to meet as a community. ELIZA’s Marcel Ackermann, Machine Learning Scientist and Scientific Coordinator for ELIZA/ELLIS, described a joint meeting with fellow Zuse Schools relAI and SECAI, where the group had the pleasure of meeting Birgit Siebe-Herbig, Head of Section Research and Internationalisation at the DAAD German Academic Exchange Service — the Projektträger (project sponsor) behind the Zuse Schools.

It was, in Ackermann’s words, “a lovely, relaxed opportunity to get to know each other better.” Siebe-Herbig shared that DAAD is genuinely pleased to see Zuse School students represented at IJCAI this year, and it was equally clear how much the students themselves valued the chance to attend a major AI conference and engage with the broader research community. The moment was shared among Elizaveta Mironovich (ELIZA), Stefani Kecman (ELIZA), Muhammad Ahmed El-Mahdy (relAI), and Birgit Siebe-Herbig (DAAD).

Research on show: from clinical reasoning to agentic data science

ELIZA students also engaged closely with the conference’s technical program. Stefani Kecman highlighted a paper by Andrea Terenziani (Polytechnic University of Turin), “Singly-Connected Multiple Minimal Networks for Efficient Temporal Reasoning About Multiple Clinical Guideline Instantiations.” The work tackles the challenge of applying multiple computer-interpretable clinical guidelines to multiple patients simultaneously, proposing a two-level framework — combining a high-level guideline representation (Asbru) with a low-level temporal constraint representation (STP) — and a new data structure, SCM-Mnet, that scales efficiently by avoiding global recomputation as the number of patients or constraints grows.

ELIZA’s support also extended to research from its network beyond the school itself: at the AutoAI-FM Workshop, part of IJCAI 2026, Kartik Nagaraj Nayak and co-authors Nitishkumar Solpure and Niladri Mitra (University of Freiburg), supervised by Tom Zehle, Alexander Pfefferle, and Omar Swelam, and guided by Professor Frank Hutter, presented “Towards Benchmarking Agentic Data Scientists.” The paper benchmarks four agentic data science systems — MLZero, AgentK, AIDE, and MLE-STAR — against the AutoGluon AutoML baseline across ten tabular machine learning datasets. The team found that AutoGluon remained a remarkably strong and consistent baseline, that agentic systems could achieve strong results on some tasks but showed higher variability and more execution failures, and that reliable execution and robustness matter just as much as raw predictive performance when evaluating autonomous ML systems. The work was supported by the University of Freiburg and Zuse School ELIZA.

Keynotes, ethics, and celebrating together

As the conference progressed, ELIZA students continued to share what they were learning. On Day 6, Elizaveta Mironovich reported on the keynote “Good Visual Representations: Easier Said Than Done” by Tinne Tuytelaars, which underscored the importance of sensible low-level representations in computer vision — a lesson she noted she would carry into her own research — alongside the “AI Ethics, Trust and Fairness” track of technical talks.

Beyond the science, the conference offered plenty of opportunities to connect with the wider community. A Student Reception at Alte Werft, near Bremen’s harbour, brought good music and a great atmosphere, while the community also gathered for an evening reception described as “such a good time,” complete with an enthusiastic DJ. As Mironovich put it, while researchers come to conferences to discuss the future of the field, all science is ultimately done by people — making it just as important and rewarding to build connections. Zuse School ELIZA’s presence at IJCAI reflected exactly that spirit, promoting networking and community within the AI research world.

Made possible by DAAD and BMBF

Zuse School ELIZA’s participation in IJCAI 2026 was funded by the DAAD German Academic Exchange Service and the Bundesministerium für Bildung und Forschung (BMBF), continuing DAAD’s role as the Projektträger for the Zuse Schools initiative. The experience gave ELIZA students the chance to learn from leading researchers, share their own work, and strengthen ties across the Zuse School network — from opening keynote to closing reception.