NAISS
SUPR
NAISS Projects
SUPR
Conformal Prediction with LLMs
Dnr:

NAISS 2025/22-1588

Type:

NAISS Small Compute

Principal Investigator:

Henrik Boström

Affiliation:

Kungliga Tekniska högskolan

Start Date:

2025-11-14

End Date:

2026-12-01

Primary Classification:

10210: Artificial Intelligence

Webpage:

Allocation

Abstract

This project will investigate the use of LLMs in conjunction with conformal prediction as implemented in the crepes Python package. Different ways of forming non-conformity scores from the generated output and embeddings of the LLMs will be investigated, and evaluated on common tasks, such as text classification.