Section quality improves with more source chunks up to k=10, beyond which gains plateau¶

Claim · All claims
Evidence · 1 study · causal r2 · q2
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person · recheck by 2029: tool and model accuracy changes quickly

Subclaims¶

q2 i? From k=1 to k=10 source chunks, overall, structure, and educational quality scores improve by +0.22, +0.16, and +0.13 on a 1–5 rubric; from 10 to 30 chunks overall improves only +0.05. → Jiawen Tao 2026

Evidence¶

Jiawen Tao 2026¶

Jiawen Tao, Miao Peng, Yaoming Li, Xiaokun Yuan, Mengzhou Wu, Wenhan Yu, Guoan Wang, Nuo Chen, Tong Yang, Maxm Pan. (2026). Beyond Rephrasing: Book-Level Organization Improves Synthetic Textbook Data for Mid-Training. arXiv preprint. https://arxiv.org/abs/2607.28109 doi:10.48550/arxiv.2607.28109

q2 · i? · causal · r2

Intrinsic-quality ablation: 400 individual sections generated with k∈{1,3,5,10,20,30} source chunks, judged by Gemini-3.1-Pro on a fixed 1–5 rubric at the section level. The article reports the printed score deltas.

"From k=1 to k=10, overall, structure, and educational scores improve by +0.22, +0.16, and +0.13. From 10 to 30 chunks, overall improves only +0.05 while breadth and depth change by <0.02 ; k=20 nearly doubles input for a +0.009 overall gain."

Discussion¶

Related Claims¶

• Preserving book-level document structure during mid-training improves downstream performance beyond identical content split into sections (+1.02 mean) — related
• Pipeline components play complementary roles: hierarchical generation has the largest effect, grounding improves specificity and factual accuracy, and cluster-informed TOC planning improves structure and audience fit — related