BeyondWeb: Lessons from Scaling Synthetic Data for Trillion-scale Pretraining
Abstract
Recent advances in large language model (LLM) pretraining have shown that simply scaling data quantity eventually leads to diminishing returns, hitting a data wall. In response, the use of synthetic data for pretraining has emerged as a promising paradigm for pushing the frontier of performance. Despite this, the factors affecting synthetic data quality remain poorly understood. In this work, we introduce BeyondWeb, a synthetic data generation framework that produces high-quality synthetic data for pretraining. BeyondWeb significantly extends the capabilities of traditional web-scale datasets, outperforming state-of-the-art synthetic pretraining datasets such as Cosmopedia and Nemotron-CC's high-quality synthetic subset (Nemotron-Synth) by up to 5.1 percentage points (pp) and 2.6pp, respectively, when averaged across a suite of 14 benchmark evaluations. It delivers up to 7.7x faster training than open web data and 2.7x faster than Nemotron-Synth. Remarkably, a 3B model trained for 180B tokens on BeyondWeb outperforms an 8B model trained for the same token budget on Cosmopedia. We also present several insights from BeyondWeb on synthetic data for pretraining: what drives its benefits, which data to rephrase and how, and the impact of model size and family on data quality. Overall, our work shows that there's no silver bullet for generating high-quality synthetic pretraining data. The best outcomes require jointly optimizing many factors, a challenging task that requires rigorous science and practical expertise. Naive approaches can yield modest improvements, potentially at great cost, while well-executed methods can yield transformative improvements, as exemplified by BeyondWeb.
Cite
@article{arxiv.2508.10975,
title = {BeyondWeb: Lessons from Scaling Synthetic Data for Trillion-scale Pretraining},
author = {DatologyAI and : and Pratyush Maini and Vineeth Dorna and Parth Doshi and Aldo Carranza and Fan Pan and Jack Urbanek and Paul Burstein and Alex Fang and Alvin Deng and Amro Abbas and Brett Larsen and Cody Blakeney and Charvi Bannur and Christina Baek and Darren Teh and David Schwab and Haakon Mongstad and Haoli Yin and Josh Wills and Kaleigh Mentzer and Luke Merrick and Ricardo Monti and Rishabh Adiga and Siddharth Joshi and Spandan Das and Zhengping Wang and Bogdan Gaza and Ari Morcos and Matthew Leavitt},
journal= {arXiv preprint arXiv:2508.10975},
year = {2025}
}
Comments
Blog version can be viewed at: http://blog.datologyai.com/beyondweb