Ayhan Suleymanzade

Hey there!

I am a Visiting Researcher at AITHYRA, supervised by Prof. Ismail Ilkan Ceylan and Prof. Michael Bronstein. My research interests lie in the intersection of Generative Modeling, Large Language Models, and AI for Science. Specifically, I am interested in continous and implicit reaosoning architectures, such as Latent Reasoning models and its applications in scientific domains (e.g. molecular dynamic, drug discovery).

I finished my Bachelor’s degree in CS/EE at KAIST, South Korea. There, I conducted research in the areas of Geometric Deep Learning, Generative Flow Networks (GFNs), and Deep Reinforcement Learning (DRL).

Also, I was lucky to be working with Prof. Stefanie Jegelka at TUM during my exchange semester.

In addition, I love solving challenging Physics problems and teaching others. I lead team Azerbaijan in International Olympiads (IPhO, EuPhO).

I am always looking for new opportunities, thus I am open to any research collaboration. Feel free to check out my CV and Transcript.

news

Nov 24, 2025	I am deeply humbled to receive the KAIST Q-Day Award 2025 [Advanced Research Division] in recognition of my contributions to academic research. 🎉
Nov 15, 2025	Our paper, Limits of Continuous Chain-of-Thought, has been accepted to AAAI 2026 Bridge on Logical and Symbolic Reasoning in Language Models 🎉
May 15, 2025	Our paper, CultDiff, has been accepted to ACL 2025 Main Conference as Oral Presentation! 🎉
Apr 16, 2025	I started my exchange semester at TU Munich, Germany. I am working with Prof. Stefanie Jegelka.
Feb 20, 2025	I am honored to be recognized on the Dean’s List for the Fall 2024 semester for the 4th time. 🎉
Feb 15, 2025	Our paper, Random Walk Neural Networks, has been accepted to ICLR 2025 as Spotlight Presentation! 🎉
Jan 10, 2025	I have started my internship at Mercari Inc, Japan.

Selected Publications

AAAI 2026 Bridge

Limits of Continuous Chain-of-Thought in Multi-Step and Multi-Chain Reasoning

Ayhan Suleymanzade^*, Andreas Bergmeister^*, and Stefanie Jegelka

AAAI 2026 Bridge on Logical and Symbolic Reasoning in Language Models, 2026

Bib PDF

@article{suleymanzade2026limits,
  title = {Limits of Continuous Chain-of-Thought in Multi-Step and Multi-Chain Reasoning},
  author = {Suleymanzade, Ayhan and Bergmeister, Andreas and Jegelka, Stefanie},
  year = {2026},
  journal = {AAAI 2026 Bridge on Logical and Symbolic Reasoning in Language Models},
}

ACL 2025 Oral

Diffusion Models Through a Global Lens: Are They Culturally Inclusive?

Zahra Bayramli^*, Ayhan Suleymanzade^*, Na Min An, and 5 more authors

Association for Computational Linguistics (ACL), 2025

Bib PDF

@article{bayramli2025diffusionmodelsgloballens,
  title = {Diffusion Models Through a Global Lens: Are They Culturally Inclusive?},
  author = {Bayramli, Zahra and Suleymanzade, Ayhan and An, Na Min and Ahmad, Huzama and Kim, Eunsu and Park, Junyeong and Thorne, James and Oh, Alice},
  year = {2025},
  journal = {Association for Computational Linguistics (ACL)},
  oral = {Oral},
}

ICLR 2025 Spotlight

Revisiting Random Walks for Learning on Graphs

Jinwoo Kim, Olga Zaghen^*, Ayhan Suleymanzade^*, and 2 more authors

International Conference on Learning Representations (ICLR), 2025

Bib PDF

@article{kim2025revisitingrandomwalkslearning,
  title = {Revisiting Random Walks for Learning on Graphs},
  author = {Kim, Jinwoo and Zaghen, Olga and Suleymanzade, Ayhan and Ryou, Youngmin and Hong, Seunghoon},
  year = {2025},
  journal = {International Conference on Learning Representations (ICLR)},
  spotlight = {Spotlight},
}

NeurIPS 2023 Spotlight

Learning Probabilistic Symmetrization for Architecture Agnostic Equivariance

Jinwoo Kim, Tien Dat Nguyen, Ayhan Suleymanzade, and 2 more authors

Neural Information Processing Systems (NeurIPS), 2023

Bib PDF

@article{kimd2024learningprobabilisticsymmetrizationarchitecture,
  title = {Learning Probabilistic Symmetrization for Architecture Agnostic Equivariance},
  author = {Kim, Jinwoo and Nguyen, Tien Dat and Suleymanzade, Ayhan and An, Hyeokjun and Hong, Seunghoon},
  year = {2023},
  journal = {Neural Information Processing Systems (NeurIPS)},
  spotlight = {Spotlight},
}