Show simple item record

Optimizing Demonstration Selection for In-Context Learning using Data Map

dc.rights.licenseCC-BY-NC-ND
dc.contributor.advisorNguyen, Dong
dc.contributor.authorChen, Silin
dc.date.accessioned2024-12-17T00:01:45Z
dc.date.available2024-12-17T00:01:45Z
dc.date.issued2024
dc.identifier.urihttps://studenttheses.uu.nl/handle/20.500.12932/48256
dc.description.sponsorshipUtrecht University
dc.language.isoEN
dc.subjectIn-context learning is a technique in which a model leverages demonstrations provided in the input context to perform tasks, without requiring parameter updates. However, existing selection methods that require the selection of a specialised demonstration set for each query impose significant computational overhead. Inspired by Data Maps (Swayamdipta et al., 2020), this thesis proposes an alternative approach to improve in-context learning by categorizing the dataset into three regions: easy-to-
dc.titleOptimizing Demonstration Selection for In-Context Learning using Data Map
dc.type.contentMaster Thesis
dc.rights.accessrightsOpen Access
dc.subject.courseuuComputing Science
dc.thesis.id41786


Files in this item

Thumbnail
Name:
Thesis of Silin Chen.pdf
Size:
3.660Mb
Format:
PDF
View/Open

This item appears in the following Collection(s)

Show simple item record