The Non-Euclidean Foundation Models and Geometric Learning Workshop will take place at NeurIPS 2025 in San Diego, CA, USA, from December 2–7, 2025. We invite you to join discussions on non-Euclidean representation learning, geometric deep learning, and large foundation models!


Workshop Summary

In the era of foundation models and Large Language Models (LLMs), Euclidean space is the de facto geometric setting of our machine learning architectures. However, recent literature has demonstrated that this choice comes with fundamental limitations. Non-Euclidean learning is quickly gaining traction. Non-Euclidean spaces, such as hyperbolic, spherical, and mixed-curvature spaces, have been shown to provide more efficient and effective representations for data with intrinsic geometric properties, like hierarchy, symmetry, and heterogeneity.

Integrating foundation models with non-Euclidean spaces has great potential to enhance their ability to capture and model the underlying structures and relationships in complex real-world data, leading to better performance, generalization, and interpretability. This workshop focuses on the intersection of Non-Euclidean representation learning and Foundation Models, exploring its potential benefits, challenges, and future directions.

Goal

The primary goal of this workshop is to provide a platform for researchers from diverse backgrounds—including geometric machine learning, representation learning, foundation models and LLMs, and domain-specific applications—to share knowledge, exchange ideas, and discuss recent advances and challenges in integrating these models with non-Euclidean spaces. By bridging the gap between these areas, we aim to unlock new opportunities for advancing AI beyond traditional Euclidean frameworks, thereby deepening our understanding of foundation models and enhancing their applications in various domains, such as natural language processing, computer vision, graph analysis, and scientific discovery.

Topics and Scope

The workshop will include submissions, talks, and poster sessions on topics related to the intersection of foundation models and non-Euclidean representation learning, including:

  • Theoretical Foundations: Generalization error, representation precision, curvature-dimension tradeoffs, geometric properties (curvature, geodesics, isometries), expressive power of non-Euclidean representations.
  • Architectures and Algorithms: Adapting existing foundation models to non-Euclidean spaces, developing new foundation models for non-Euclidean operations, and investigating non-Euclidean architectures in conjunction with foundation models.
  • Applications: Graph analysis, text processing, image understanding, biomedical research, and AI for scientific discovery (drug discovery, material science, climate modeling).
  • Trustworthiness and Robustness: Adversarial robustness, fairness, interpretability, and privacy in non-Euclidean foundation models.
  • Benchmarks and Tools: New datasets, evaluation protocols, and software libraries supporting the integration of foundation models and non-Euclidean representations.

Submissions and Timeline

For detailed submission guidelines, topics, and important dates, please visit our Call for Papers page.

We manage paper submissions through OpenReview. The review process is double‑blind, so submissions must be anonymized. We welcome work that is (1) original and unpublished, (2) recently published, or (3) work‑in‑progress. By default, submissions will not have archival proceedings.

Please use the NeurIPS 2025 LaTeX style file; it includes a preprint option for non‑anonymous preprints posted online (see additional formatting details here). Submissions should be PDFs of ≤ 9 pages (excluding references and appendices).

We will select outstanding papers for lightning talks. The award for best paper will be announced at the workshop.

Community Building and Discussion Opportunities

We plan to encourage interactive discussions throughout the workshop through several structured formats. During the poster session, we will organize posters based on research directions into thematic groups, facilitating focused discussions and cross-pollination of ideas. For the panel session, we will implement a multi-channel approach to question collection, including author-submitted questions and social media engagement.

Tentative Schedule

  • 8:30–8:50 AM: Poster setup
  • 8:50–9:00 AM: Opening remarks
  • 9:00–9:50 AM: Invited talk: Philip S. Yu
  • 9:50–10:40 AM: Invited talk: Pascal Mettes
  • 10:40–11:30 AM: Contributed talks
  • 11:30–11:50 AM: Discussions and coffee break
  • 11:50–12:55 PM: Poster session
  • 12:55–2:00 PM: Lunch break
  • 2:00–2:50 PM: Invited talk: Min Zhou
  • 2:50–3:40 PM: Invited talk: Bo Xiong
  • 3:40–4:30 PM: Invited talk: Smita Krishnaswamy
  • 4:30–5:00 PM: Discussions and coffee break
  • 5:00–5:30 PM: Panel discussions: Foundation Models Meet Non-Euclidean Space

Organizers

  • Menglin Yang (HKUST(GZ))
  • Neil He (UIUC)
  • Yifei Zhang (NTU)
  • Rex Ying (Yale University)

Contact

Feel free to contact us at negel2025@outlook.com. We look forward to your participation at NeurIPS 2025!