MUX selected as a spotlight at NeurIPS 2026
MUX: Continuous Reasoning via Multiplexed Tokens was accepted to NeurIPS 2026 as a spotlight paper.
News archive
News posts grouped by year.
2026 13
October 4
Saka Research has a new website
Our group website is live, with pages for our research, team, publications, and news.
New preprint: Global Transport Couplings for Classifier-Free Guided Flows
A class-agnostic optimal-transport coupling for conditional flow models that consistently improves generation when combined with classifier-free guidance.
New preprint: To Learn is to Wander
We released Wander, a single graph foundation model for node classification, link prediction, and knowledge-graph reasoning built on random walks.
September 2
RelAgent accepted to NeurIPS 2026
Our paper RelAgent: LLM Agents as Data Scientists for Relational Learning was accepted to NeurIPS 2026.
New preprint: Thinking with Looped Flows
Looped flows train recurrent models with local denoising objectives, so that extra computation at inference time can help solve harder problems.
July 1
Talks on graph foundation models
İsmail gave talks and guest lectures on graph foundation models at the LOGML summer school, EPFL, the Wallenberg Advanced Scientific Forum, and Oxford.
May 4
New preprint: What are the Right Symmetries for Formal Theorem Proving?
A category-theoretic framework for the symmetries that LLM-based theorem provers should respect.
New preprint: MUX, continuous reasoning via multiplexed tokens
MUX compresses step-by-step reasoning of language models into continuous latent tokens that each multiplex a span of reasoning.
New preprint: RelAgent
RelAgent uses LLM agents as data scientists for relational learning, producing predictors that are fast and interpretable.
Two papers accepted to ICML 2026
Categorical Flow Maps and MacroGuide: Topological Guidance for Macrocycle Generation were accepted to ICML 2026.
February 1
Louis Tichelman and Angelo Gnazzo join the group
Louis Tichelman and Angelo Gnazzo joined the group as PhD students at TU Wien.
January 1
Three papers accepted to ICLR 2026
Topological Flow Matching, HYPER, and Flock were accepted to ICLR 2026.
2025 8
December 1
İsmail is now Associate Professor at TU Wien
İsmail İlkan Ceylan was appointed Associate Professor at TU Wien.
October 1
Paper accepted to LoG 2025
One Model, Any Conjunctive Query: Graph Neural Networks for Answering Queries over Incomplete Knowledge Graphs was accepted to the Learning on Graphs Conference 2025.
September 2
Two papers accepted to NeurIPS 2025
Equivariance Everywhere All At Once and Curly Flow Matching were accepted to NeurIPS 2025.
New preprint: Bringing Graphs to the Table
Node classification reformulated as a tabular problem, so that tabular foundation models can classify nodes zero-shot.
August 1
The group moves to Vienna
İsmail joined TU Wien as Assistant Professor and AITHYRA as Adjunct PI, while remaining an Academic Member at the University of Oxford.
June 1
Lecturing at the Nepal AI School
İsmail lectured on geometric deep learning with Michael Bronstein at the Sixth Annual Nepal AI School in Kathmandu.
May 1
Paper accepted to ICML 2025
How Expressive are Knowledge Graph Foundation Models? was accepted to ICML 2025.
January 1
Paper accepted to ICLR 2025
Homomorphism Counts as Structural Encodings for Graph Learning was accepted to ICLR 2025.
2024 2
October 1
Gibbs Prize for Krzysztof Olejniczak
Krzysztof Olejniczak's MSc thesis on answering conjunctive queries with graph neural networks was awarded the Gibbs Prize.
June 1
Teaching commendation at Oxford
İsmail received a Department Teaching Commendation at the University of Oxford for the Graph Representation Learning course.
2023 1
October 1
MSc prizes for Oscar Davis and Zeyang Zhao
Oscar Davis received the Hoare Project Prize and Zeyang Zhao the Gibbs Prize for their MSc theses in the group.