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¶

Claim · All claims
Evidence · 1 study (4 entries) · 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? Removing TOC planning, clustering, and source chunks (Baseline C) drops overall judged quality from 9.38 to 6.12 (−3.26), the largest observed component effect. → Jiawen Tao 2026 q2 i? Source chunks primarily improve specificity (+0.23), examples (+0.35), and factual accuracy (+0.32). → Jiawen Tao 2026 (2) q2 i? Cluster-informed TOC planning improves audience/style fit (+0.59), structure (+0.44), pedagogy (+0.23), and overall quality (+0.14) over discipline-name-only planning. → Jiawen Tao 2026 (3) q2 i? The TOC quality gate acts as a safety net: books from failed TOCs score lower on overall (−0.25), structure (−0.20), and pedagogy (−0.19), a smaller degradation than removing source chunks or cluster summaries. → Jiawen Tao 2026 (4)

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

Component ablation on 100 matched book instances scored by Gemini-3.1-Pro on a 1–10 rubric; Baseline C generates an entire book without TOC planning, clustering, or source chunks.

"Baseline C scores 6.12 overall versus 9.38 for Full (−3.26). Despite a∼30–40K-word length target, it still yields only∼56K characters (roughly 4× shorter than Full), with lower depth (5.86) and specificity (6.41)."

Jiawen Tao 2026 (2)¶

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

q2 · i? · causal · r2

Same judged ablation; Baseline A keeps Full's cluster-informed TOC but removes source chunks, isolating retrieval grounding's contribution to domain-specific detail.

"Full versus Baseline A isolates retrieval grounding: source chunks improve specificity (+0.23), examples (+0.35), and factual accuracy (+0.32), with smaller differences elsewhere."

Jiawen Tao 2026 (3)¶

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

q2 · i? · causal · r2

Same judged ablation comparing Baseline A (cluster-informed TOC) against Baseline B (TOC from discipline name alone), neither using source chunks.

"A improves audience/style fit (9.21 vs. 8.62, +0.59), structure (9.35 vs. 8.91, +0.44), pedagogy (9.23 vs. 9.00, +0.23), and overall quality (9.21 vs. 9.07, +0.14), showing that source coverage supports more targeted TOCs before any section text is generated."

Jiawen Tao 2026 (4)¶

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

q2 · i? · causal · r2

Baseline D generates books from 200 failed TOCs, evaluated with the same judge and rubric as the other ablation conditions.

"Relative to Full, they score lower on overall (−0.25), structure (−0.20), and pedagogy (−0.19). This smaller degradation than removing source chunks or cluster summaries makes the gate a safety net rather than the primary quality driver."

Discussion¶

Related Claims¶

• Preserving book-level document structure during mid-training improves downstream performance beyond identical content split into sections (+1.02 mean) — related
• Structured book synthesis outperforms matched independent document rephrasing (+1.17 mean), while rephrasing alone ties natural books — related
• Section quality improves with more source chunks up to k=10, beyond which gains plateau — related
• Replacing natural books with synthetic textbooks improves the 28-benchmark mean by +1.09, with gains spanning all four categories — related
• Removing the cognitive load module causes the largest ablation performance drop, while state fusion has a smaller effect — related
• Ablating expert-informed structural components lowered LLM-judge scores on all dimensions, with largest drops in challenge quality, intent alignment, and domain grounding — related
• Static Knowledge Grounding and Dynamic Personal Memory are complementary: removing both yields the largest quality degradation — related
• Knowledge and behavior profile components provide complementary signals: removing either hurts metrics tied to the other — related
• Federated aggregation is the primary driver of performance gains, contributing an average 11.99% improvement — related
• NCERT-tailored RAG preprocessing and figure-aware indexing improve retrieval precision over a generic-chunking baseline, per the authors' assertion — related
• Each pedagogical verification component contributes to a distinct quality dimension: layout verification most strongly — related
• Ablation study shows metadata-enriched embeddings produce the largest accuracy drop, while attention masking, KV caching and dynamic batching primarily support inference efficiency — related