How Abilities in Large Language Models are Affected by Supervised Fine-tuning Data Composition
How Abilities in Large Language Models are Affected by Supervised Fine-tuning Data Composition
About this item
Full title
Author / Creator
Publisher
Ithaca: Cornell University Library, arXiv.org
Journal title
Language
English
Formats
Publication information
Publisher
Ithaca: Cornell University Library, arXiv.org
Subjects
More information
Scope and Contents
Contents
Large language models (LLMs) with enormous pre-training tokens and parameters emerge diverse abilities, including math reasoning, code generation, and instruction following. These abilities are further enhanced by supervised fine-tuning (SFT). While the open-source community has explored ad-hoc SFT for enhancing individual capabilities, proprietary...
Alternative Titles
Full title
How Abilities in Large Language Models are Affected by Supervised Fine-tuning Data Composition
Authors, Artists and Contributors
Identifiers
Primary Identifiers
Record Identifier
TN_cdi_proquest_journals_2885380803
Permalink
https://devfeature-collection.sl.nsw.gov.au/record/TN_cdi_proquest_journals_2885380803
Other Identifiers
E-ISSN
2331-8422
