arXiv:2608. 25826v1 Announce Type: cross Abstract: A recent line of synthetic-data work reconstructs the thinking behind existing text rather than rewriting the text itself, but it operates on short web passages, recovers only local thoughts, and leaves the structure of whole documents untouched.
A recent line of synthetic-data work reconstructs the thinking behind existing text rather than rewriting the text itself, but it operates on short web passages, recovers only local thoughts, and leav...
ConvergeWriter introduces a bottom‑up, data‑driven framework for long‑form document generation that first retrieves exhaustive knowledge from a source corpus and clusters it into distinct knowledge groups. These clusters then guide the creation of a hierarchical outline and the final text, ensuring the output is strictly grounded in the retrieved material and traceable to its sources. Experiments on 14B and 32B LLMs show that this approach matches or surpasses state‑of‑the‑art baselines, especially in scenarios requiring high factual fidelity and structural coherence.
arXiv:2604. 13977v2 Announce Type: replace-cross Abstract: Synthetic data is a standard component in training large language models, yet systematic comparisons across design dimensions, including rephrasing strategy, generator model, and source data, remain absent.
arXiv:2601. 22146v2 Announce Type: replace-cross Abstract: Due to limited supervised training data, large language models (LLMs) are typically pre-trained via a self-supervised "predict the next word" objective on a vast amount of unstructured text data.
AFCS (Atomic‑Fact‑Conditioned Synthesis) is a new pipeline for generating synthetic data for language‑model pretraining that focuses on extracting and using fine‑grained atomic facts from source documents. Unlike previous methods such as WRAP and Nemotron‑CC, AFCS explicitly requires secondary entities, list/table entries, and precise temporal attributes to appear in the synthesized corpus, employing structure‑aware extraction, source‑grounded verification, local context expansion, and multi‑format generation. In experiments with OLMo‑3‑7B and a 32B model on the FineWiki corpus, AFCS outperforms the baselines on several QA benchmarks and improves general benchmark performance by +1.09 when integrated mid‑training with 100B tokens.
arXiv:2609.37891v1 Announce Type: cross Abstract: Current pre-training datasets are derived from web crawls, with all their issues, and were not designed to support mid- and post-training pipelines--...
The paper presents a benchmark that compares seven long‑form generation frameworks across three granularities—single chapter, multi‑chapter, and whole book—using an anchor‑based LLM‑as‑a‑judge protocol to evaluate outlines directly. Results show no single framework dominates across all settings; performance depends on how well a framework’s output form matches the target granularity, with SuperWriter excelling in length‑constrained single‑chapter mode but losing advantage in whole‑book mode. The study finds only moderate correlation between outline and writing quality, supporting the idea that these two stages should be evaluated separately.
SHELF is a Python system that creates synthetic, controlled benchmark data for evaluating large language models on bibliographic tasks such as classification, clustering, retrieval, pair classification, and instruction retrieval. It generates 62,899 model-written documents based on Library of Congress vocabularies and compares methods like TF, TF‑IDF, BM25, popular encoders, and zero‑shot decoders, reporting performance metrics such as 0.8887 for subject classification and 0.2605 for genre‑form classification. The tool also allows independent variation of bibliographic facets and can produce unseen documents beyond a model’s training cutoff, with results indicating that model rankings transfer more reliably than absolute scores when compared to other benchmarks.
The paper introduces SynPro, a synthetic data generation framework that augments limited organic text for large language model pretraining by applying rephrasing and reformatting operations. SynPro’s generators are optimized with reinforcement learning rewards for quality, faithfulness, and data influence, and are updated continuously as training plateaus. Experiments on 400M, 1.1B, and 2B models show that SynPro can unlock 3.4–5.2× the effective tokens of standard repetition, even outperforming a non‑data‑bound oracle at larger scales.
The paper introduces STAIR, a retrieval system that uses a document’s Table of Contents to guide large language models in accessing global structure, thereby reducing hallucinations in Retrieval Augmented Generation. Experiments with a fine‑tuned Differentiable Search Index show that ToC‑based retrieval yields a low hallucination rate (<0.05%) and improves Recall@1 to 82.6% on the newly released SearchTome benchmark, outperforming baselines like BM25, DPR, and Mistral. The authors also release SearchTome, a diverse dataset of 18 books across six domains, to encourage further research in ToC‑based retrieval.
arXiv:2608.22817v1 Announce Type: new Abstract: Industrial technical reports contain high-value knowledge for maintenance, troubleshooting, and product engineering, but their heterogeneous structure...