Metraipolis › archive › dc_445179d45cca
AFCS: Atomic-Fact-Conditioned Synthesis for Factual Knowledge Pretraining — arXiv:2610.20086
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2610.20086
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-15T20:02:54.627Z
- SHA-256 of the archived text
db7a311b3a79ef16ae1d438083b9e1f1ed9e972dce9a2cdc07959f10e9756a61- Stable URL of this page
https://metraipolis.com/documento/dc_445179d45cca
How to cite
AFCS: Atomic-Fact-Conditioned Synthesis for Factual Knowledge Pretraining — arXiv:2610.20086 — arXiv (en), https://arxiv.org/abs/2610.20086 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_445179d45cca (Metraipolis archive dc_445179d45cca, sha256 db7a311b3a79ef16ae1d438083b9e1f1ed9e972dce9a2cdc07959f10e9756a61)
BibTeX
@misc{dc445179d45cca,
title = {AFCS: Atomic-Fact-Conditioned Synthesis for Factual Knowledge Pretraining — arXiv:2610.20086},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2610.20086},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_445179d45cca; sha256 db7a311b3a79ef16ae1d438083b9e1f1ed9e972dce9a2cdc07959f10e9756a61},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.20086. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Synthetic data is increasingly used in language-model pretraining. Existing knowledge-corpus synthesis pipelines, including WRAP and Nemotron-CC, usually take a source passage or document segment as input and generate rewritten or QA-style samples from that text. This can improve surface form and prompt diversity, but it does not explicitly require fine-grained source facts -- especially secondary entities, list/table entries, and precise temporal attributes -- to appear in the synthesized corpus. We propose AFCS (Atomic-Fact-Conditioned Synthesis), a pipeline that first extracts source-supported atomic facts and then uses each retained fact to drive generation. AFCS uses structure-aware extraction over the document tree, source-grounded verification, local context expansion, and multi-format generation. In continued-pretraining experiments with OLMo-3-7B and a 32B base on the same FineWiki source corpus, AFCS outperforms WRAP and Nemotron-CC on SimpleQA, SimpleQA Verified, PopQA, and HotpotQA. Granularity ablations show that fact-level synthesis improves over chunk-level, paragraph-level, and entity-level synthesis; and a 100B-token mid-training integration improves factual benchmarks and raises the 12-task general benchmark average by +1.09. Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno). Categorie arXiv: Computation and Language (cs.CL) ; Artificial Intelligence (cs.AI). Pagina del lavoro: https://arxiv.org/abs/2610.20086 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito). Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina. Letto il 2026-10-15T19:35:48.400Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).