Project Details
Description
At ETH Zürich I will be working on understanding the fundamental
mathematical structures in language that allow for matrix models (such as
transformers on which ChatGPT is based) to be as succesful in modeling
language as they are. We use category theory to study embedding spaces and specifically a notion of logical implication. Standard approaches to embeddings have succeeded in unveiling some structure in language, but viewing language through the lens of category theory could unveil others.
mathematical structures in language that allow for matrix models (such as
transformers on which ChatGPT is based) to be as succesful in modeling
language as they are. We use category theory to study embedding spaces and specifically a notion of logical implication. Standard approaches to embeddings have succeeded in unveiling some structure in language, but viewing language through the lens of category theory could unveil others.
| Status | Finished |
|---|---|
| Effective start/end date | 01/09/2025 → 31/12/2025 |
Collaborative partners
- IT University of Copenhagen (lead)
- Swiss Federal Institute of Technology Zürich
Funding
- Stibo-Fonden: DKK50,000.00
Fingerprint
Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.