_version_ 1866915417333694464
author Gohari, Hajar Emami
Kadhe, Swanand Ravindra
Shah, Syed Yousaf
Adam, Constantin
Adebayo, Abdulhamid
Adusumilli, Praneet
Ahmed, Farhan
Angel, Nathalie Baracaldo
Borse, Santosh Subhashrao
Chang, Yuan-Chi
Dang, Xuan-Hong
Desai, Nirmit
Eres, Revital
Iwamoto, Ran
Karve, Alexei
Koyfman, Yan
Lee, Wei-Han
Liu, Changchang
Lublinsky, Boris
Ohko, Takuyo
Pesce, Pablo
Touma, Maroun
Wang, Shiqiang
Witherspoon, Shalisha
Woisetschläger, Herbert
Wood, David
Wu, Kun-Lung
Yoshida, Issei
Zawad, Syed
Zerfos, Petros
Zhou, Yi
Bhattacharjee, Bishwaranjan
author_facet Gohari, Hajar Emami
Kadhe, Swanand Ravindra
Shah, Syed Yousaf
Adam, Constantin
Adebayo, Abdulhamid
Adusumilli, Praneet
Ahmed, Farhan
Angel, Nathalie Baracaldo
Borse, Santosh Subhashrao
Chang, Yuan-Chi
Dang, Xuan-Hong
Desai, Nirmit
Eres, Revital
Iwamoto, Ran
Karve, Alexei
Koyfman, Yan
Lee, Wei-Han
Liu, Changchang
Lublinsky, Boris
Ohko, Takuyo
Pesce, Pablo
Touma, Maroun
Wang, Shiqiang
Witherspoon, Shalisha
Woisetschläger, Herbert
Wood, David
Wu, Kun-Lung
Yoshida, Issei
Zawad, Syed
Zerfos, Petros
Zhou, Yi
Bhattacharjee, Bishwaranjan
contents Data quantity and quality play a vital role in determining the performance of Large Language Models (LLMs). High-quality data, in particular, can significantly boost the LLM's ability to generalize on a wide range of downstream tasks. Large pre-training datasets for leading LLMs remain inaccessible to the public, whereas many open datasets are small in size (less than 5 trillion tokens), limiting their suitability for training large models. In this paper, we introduce GneissWeb, a large dataset yielding around 10 trillion tokens that caters to the data quality and quantity requirements of training LLMs. Our GneissWeb recipe that produced the dataset consists of sharded exact sub-string deduplication and a judiciously constructed ensemble of quality filters. GneissWeb achieves a favorable trade-off between data quality and quantity, producing models that outperform models trained on state-of-the-art open large datasets (5+ trillion tokens). We show that models trained using GneissWeb dataset outperform those trained on FineWeb-V1.1.0 by 2.73 percentage points in terms of average score computed on a set of 11 commonly used benchmarks (both zero-shot and few-shot) for pre-training dataset evaluation. When the evaluation set is extended to 20 benchmarks (both zero-shot and few-shot), models trained using GneissWeb still achieve a 1.75 percentage points advantage over those trained on FineWeb-V1.1.0.
format Preprint
id arxiv_https___arxiv_org_abs_2502_14907
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle GneissWeb: Preparing High Quality Data for LLMs at Scale
Gohari, Hajar Emami
Kadhe, Swanand Ravindra
Shah, Syed Yousaf
Adam, Constantin
Adebayo, Abdulhamid
Adusumilli, Praneet
Ahmed, Farhan
Angel, Nathalie Baracaldo
Borse, Santosh Subhashrao
Chang, Yuan-Chi
Dang, Xuan-Hong
Desai, Nirmit
Eres, Revital
Iwamoto, Ran
Karve, Alexei
Koyfman, Yan
Lee, Wei-Han
Liu, Changchang
Lublinsky, Boris
Ohko, Takuyo
Pesce, Pablo
Touma, Maroun
Wang, Shiqiang
Witherspoon, Shalisha
Woisetschläger, Herbert
Wood, David
Wu, Kun-Lung
Yoshida, Issei
Zawad, Syed
Zerfos, Petros
Zhou, Yi
Bhattacharjee, Bishwaranjan
Computation and Language
Artificial Intelligence
Data quantity and quality play a vital role in determining the performance of Large Language Models (LLMs). High-quality data, in particular, can significantly boost the LLM's ability to generalize on a wide range of downstream tasks. Large pre-training datasets for leading LLMs remain inaccessible to the public, whereas many open datasets are small in size (less than 5 trillion tokens), limiting their suitability for training large models. In this paper, we introduce GneissWeb, a large dataset yielding around 10 trillion tokens that caters to the data quality and quantity requirements of training LLMs. Our GneissWeb recipe that produced the dataset consists of sharded exact sub-string deduplication and a judiciously constructed ensemble of quality filters. GneissWeb achieves a favorable trade-off between data quality and quantity, producing models that outperform models trained on state-of-the-art open large datasets (5+ trillion tokens). We show that models trained using GneissWeb dataset outperform those trained on FineWeb-V1.1.0 by 2.73 percentage points in terms of average score computed on a set of 11 commonly used benchmarks (both zero-shot and few-shot) for pre-training dataset evaluation. When the evaluation set is extended to 20 benchmarks (both zero-shot and few-shot), models trained using GneissWeb still achieve a 1.75 percentage points advantage over those trained on FineWeb-V1.1.0.
title GneissWeb: Preparing High Quality Data for LLMs at Scale
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2502.14907