Pith. sign in

Synthetic bootstrapped pretraining , url =

Yang, Zitong, Zhang, Aonan, Liu, Hong, Hashimoto, Tatsunori, Candès, Emmanuel, Wang, Chong · arXiv (Cornell University) · DOI 10.48550/arxiv.2509.15248 · arXiv 2509.15248

1 Pith paper cite this work. Polarity classification is still indexing.

1 Pith paper citing it
open at publisher browse 1 citing papers arXiv PDF

fields

cs.LG 1

years

2026 1

verdicts

CONDITIONAL 1

representative citing papers

Bridging Compute- and Data-Optimal Pretraining

cs.LG · 2026-07-28 · conditional · novelty 7.0

Pretraining loss obeys a single law in which repeated or paraphrased tokens count as η(N, data-per-parameter, expansion-ratio) fresh tokens, with total effective data saturating as derived tokens grow.

citing papers explorer

Showing 1 of 1 citing paper.

  • Bridging Compute- and Data-Optimal Pretraining cs.LG · 2026-07-28 · conditional · none · ref 53

    Pretraining loss obeys a single law in which repeated or paraphrased tokens count as η(N, data-per-parameter, expansion-ratio) fresh tokens, with total effective data saturating as derived tokens grow.