Morand, M Victor (2024) On the representations of entities in Auto-regressive Large Language Models PFE - Project Graduation, ENSTA.

[img]
Preview
PDF
1335Kb

Abstract

This document is a technical report that aims at presenting the work I’ve been doing at ISIR during my end-of-studies Internship, as part of the ENSTA degree. From April to October 2024, I have been working in the MLIA team on the topic of entity representations in recent large language models. This work is in line with the trend towards explainable and responsible large language models for knowledge management. There is indeed a growing effort to explore ways to explain and manage how large language model stores and retrieves factual information in what it reads and trains on. My internship and the following PhD aim at searching in that direction. I am very proud to say that the work that I have been doing in the lab for the last six month have led to the redaction of a conference paper that will be submitted to the annual conference of the North American Chapter of the Association for Computational Linguistics. The main body of this document, namely sections 6 to Section 15, as well as the appendix, are made from the contend of the paper, which is still a work in progress.

Item Type:Thesis (PFE - Project Graduation)
Uncontrolled Keywords:Natural Language Processing
Subjects:Information and Communication Sciences and Technologies
Mathematics and Applications
ID Code:10413
Deposited By:Victor MORAND
Deposited On:08 oct. 2024 15:48
Dernière modification:08 oct. 2024 15:48

Repository Staff Only: item control page