Speaker
Description
Foundation models are a generalized type of large language model (LLM) that are trained on large, diverse, unlabeled datasets and can perform a variety of downstream tasks. The success of LLMs in industry has motivated research into their application in scientific domains. High energy nuclear and particle physics experiments, such as sPHENIX, produce vast amounts of sparse, low-level data that could be trained to perform reconstruction tasks. In this talk we will discuss recent work on the application of a foundational model to sPHENIX Time Projection Chamber (TPC) data [arXiv:2508.14087] and its performance with several TPC-related downstream tasks. The implications of Foundation Models, including additional challenges that they could address, will be highlighted.
| Is this an experimental talk? | Yes |
|---|---|
| Is this on behalf of a collaboration? | Yes |
| Which collaboration? | sPHENIX Collaboration |
| Are you willing to present as a poster if it is not selected for oral presentation? | Yes |