aiqbench analysis
This research introduces SPOKES, a probabilistic framework that directly optimizes dataset diversity during pretraining to reduce redundancy without relying on imperfect proxies. By jointly maximizing quality and variety using the G-Vendi score, the method achieves significant downstream performance gains over standard sampling or existing deduplication techniques on major corpora like FineWeb and DCLM. For practitioners managing limited compute budgets, this approach offers a rigorous mathematical pathway to construct more efficient and robust foundation models.
Spokes: Optimizing for Diverse Pretraining Data Selection
arXiv:2606.15216v1 Announce Type: new Abstract: Diversity plays a critical role in data selection, improving performance under fixed data budgets by reducing redundancy and repetition. However, optimizing for diversity is inherently challenging, as it is a set-level property that depends on interactions between data points rather than individual examples. As a result, existing approaches typically rely on proxies or approximations, which often fail to ensure sufficiently diverse subsets. In this work, we directly optimize diversity by introducing a probabilistic diversification framework based on the G-Vendi score, optimized via exponentiated gradient descent. Our method produces subsets that are substantially more diverse than those obtained via random sampling, achieving a +489 increase in G-Vendi score on a 500k-sample subset. We evaluate our approach on FineWeb and DCLM, where it consistently outperforms existing methods. Notably, SPOKES (diversity-only) improves average downstream performance by +0.4 and +0.5 points over random sampling on DCLM and FineWeb, respectively. More importantly, jointly optimizing for both quality and diversity yields the strongest results: SPOKES achieves gains of +1.5 and +1.4 points on DCLM and FineWeb, outperforming all baselines, including semantic deduplication and quality filtering.
Abstract:Diversity plays a critical role in data selection, improving performance under fixed data budgets by reducing redundancy and repetition. However, optimizing for diversity is inherently challenging, as it is a set-level property that depends on interactions between data points rather than individual examples. As a result, existing approaches typically rely on proxies or approximations, which often fail to ensure sufficiently diverse subsets. In this work, we directly optimize diversity by introducing a probabilistic diversification framework based on the G-Vendi score, optimized via exponentiated gradient descent. Our method produces subsets that are substantially more diverse than those obtained via random sampling, achieving a +489 increase in G-Vendi score on a 500k-sample subset. We evaluate our approach on FineWeb and DCLM, where it consistently outperforms existing methods. Notably, SPOKES (diversity-only) improves average downstream performance by +0.4 and +0.5 points over random sampling on DCLM and FineWeb, respectively. More importantly, jointly optimizing for both quality and diversity yields the strongest results: SPOKES achieves gains of +1.5 and +1.4 points on DCLM and FineWeb, outperforming all baselines, including semantic deduplication and quality filtering.
Computation and Language (cs.CL); Artificial Intelligence (cs.AI)
arXiv:2606.15216 [cs.CL]
(or arXiv:2606.15216v1 [cs.CL] for this version)
https://doi.org/10.48550/arXiv.2606.15216 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 277 words
Related coverage
Towards Scalable Data Diversification for Language Model Pretraining via Leverage Score Sampling
arXiv · Sep 29, 2026
Reflection’s Beam model hits 501 billion parameters
Quantum Zeitgeist · Oct 5, 2026
Introducing Instella-MoE: A State-of-the-Art Fully Open Mixture-of-Experts Language Model
AMD · Jul 24, 2026
[AINews] The Future of Latent Space
Latent Space · Sep 25, 2026