Quest: Query-centric Data Synthesis Approach for Long-context Scaling of Large Language Model

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Gao, Chaochen, Wu, Xing, Fu, Qi, Hu, Songlin
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866929708347686912
author Gao, Chaochen
Wu, Xing
Fu, Qi
Hu, Songlin
author_facet Gao, Chaochen
Wu, Xing
Fu, Qi
Hu, Songlin
contents Recent advancements in large language models (LLMs) have highlighted the importance of extending context lengths for handling complex tasks. While traditional methods for training on long contexts often use filtered long documents, these approaches lead to domain imbalances, limiting model performance. To address this, techniques like random document concatenation (Standard) and similarity-based methods (KNN, ICLM) have been developed. However, they either sacrifice semantic coherence or diversity. To balance both aspects, we introduce Quest, a query-centric data synthesis method aggregating semantically relevant yet diverse documents. Quest uses a generative model to predict potential queries for each document, grouping documents with similar queries and keywords. Extensive experiments demonstrate Quest's superior performance on long-context tasks, achieving remarkable results with context lengths of up to 1M tokens and confirming its scalability across various model sizes.
format Preprint
id arxiv_https___arxiv_org_abs_2405_19846
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Quest: Query-centric Data Synthesis Approach for Long-context Scaling of Large Language Model
Gao, Chaochen
Wu, Xing
Fu, Qi
Hu, Songlin
Computation and Language
Artificial Intelligence
Recent advancements in large language models (LLMs) have highlighted the importance of extending context lengths for handling complex tasks. While traditional methods for training on long contexts often use filtered long documents, these approaches lead to domain imbalances, limiting model performance. To address this, techniques like random document concatenation (Standard) and similarity-based methods (KNN, ICLM) have been developed. However, they either sacrifice semantic coherence or diversity. To balance both aspects, we introduce Quest, a query-centric data synthesis method aggregating semantically relevant yet diverse documents. Quest uses a generative model to predict potential queries for each document, grouping documents with similar queries and keywords. Extensive experiments demonstrate Quest's superior performance on long-context tasks, achieving remarkable results with context lengths of up to 1M tokens and confirming its scalability across various model sizes.
title Quest: Query-centric Data Synthesis Approach for Long-context Scaling of Large Language Model
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2405.19846