NEGEL Workshop @ TheWebConf 2025

The Workshop on Non-Euclidean Foundation Models and Geometric Learning (NEGEL) will take place at TheWebConf 2025 in Sydney, Australia, from April 28–May 2, 2025. We invite you to join discussions on Non-Euclidean representation learning and geometric deep learning, and large foundation models, alongside web-related applications!

The primary objective of the NEGEL workshop at TheWebConf 2025 is to create a collaborative platform for researchers, industry professionals, and academics from diverse fields. These include non-Euclidean representation learning, geometric deep learning, and large foundation models, as well as web-related applications such as recommender systems, social network analysis, web knowledge graphs, search systems, information retrieval, and multi-modal web content understanding. The workshop aims to foster knowledge sharing, idea exchange, and discussions on the latest advancements and challenges in these domains.

Important Dates

2025-02-01: Paper submission deadline(Extended)
2025-02-03: Author notification (Extended)
2025-02-07: Camera-ready submission
2025-04-28: Workshop at TheWebConf 2025 (Whole-day session)
Submission site: https://openreview.net/group?id=ACM.org/TheWebConf/2025/Workshop/NEGEL
Venue: International Convention & Exhibition Centre (ICC), Sydney, Australia
Timezone: Anywhere On Earth (UTC-12)

Program Highlights

Interactive Poster Sessions: Two dynamic sessions to present and discuss the latest research in Geometric Learning, non-Euclidean representation learning and foundation models, along with web-related applications
Contributed Talks: Selected high-quality papers presented in two dedicated slots, with approximately 4 papers chosen for oral presentations
Panel Discussion: “FM & Geometric Non-Euclidean Learning” exploring future directions and challenges in combining foundation models (e.g., LLMs) with non-Euclidean approaches
Networking Opportunities:
- Research matchmaking during coffee breaks
- Informal discussions during lunch
- Breakout discussion sessions
- Opening ice-breaking activities
Awards:
- Best Paper Award for exceptional contributions
- Best Poster Presentation Award for effective presentation and engagement
- Travel awards to support students and authors needing financial assistance

Call for Papers

We welcome submissions on the following topics (but not limited to):

New Large Language/Foundation Models with Geometries: Develop new architectures and algorithms integrating Large Language/Foundation Models (e.g., LLMs, ViTs, and multi-modal models) with geometric and non-Euclidean learning, enhancing their capability to handle non-Euclidean structures, complex relationships understanding, and complex reasoning abilities.
Large Language/Foundation Models Adaptation with Geometries: Present methodologies for extending and adapting large foundation/language models through geometric guidances. Key areas include geometric fine-tuning strategies, manifold-aware continual learning, and techniques for preserving geometric structure during model adaptation. Special interest in GraphRAG implementations that leverage geometric structures for knowledge retrieval and generation, as well as geometric approaches to agent architectures.
Theoretical Studies: Investigate fundamental principles of geometric and non-Euclidean space learning, including manifold theory, differential geometry, and non-Euclidean spaces. Topics encompass mathematical properties such as curvature, geodesic distances, parallel transport, and isometric transformations, with particular emphasis on their implications for foundation model architectures and optimization dynamics.
Geometric Deep Learning: Advanced methods in geometric learning, including graph neural networks, equivariant models, and message-passing frameworks, with applications to web-specific tasks like structured data representation and multi-modal alignment.
Web Applications and Case Studies: Highlight applications in web domains, including social networks, knowledge graphs, search systems, recommender systems, and multi-modal content understanding.
Trustworthiness and Robustness: Tackle challenges of fairness, interpretability, adversarial robustness, and privacy in non-Euclidean and geometric models, especially for user-facing web applications.
Benchmarks and Evaluation: Propose new datasets, benchmarks, and evaluation protocols, emphasizing scalability and real-world performance on non-Euclidean and geometric models.