Large-Scale Diverse Synthesis for Mid-Training

• URL: http://arxiv.org/abs/2508.01326v1
• Date: Sat, 02 Aug 2025 11:37:16 GMT
• Title: Large-Scale Diverse Synthesis for Mid-Training
• Authors: Xuemiao Zhang, Chengying Tu, Can Ren, Rongxiang Weng, Hongfei Yan, Jingang Wang, Xunliang Cai,
• Abstract summary: BoostQA is a 100B-token large-scale question-answering dataset.<n>We propose a novel diversified pipeline to synthesize BoostQA.<n>Our method enables Llama-3 8B, mid-trained on a 40B-token dataset, to achieve an average improvement of $mathbf12.74%$ on MMLU and CMMLU.
• Score: 15.81154701009597
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: The scarcity of high-quality, knowledge-intensive training data hinders the development of large language models (LLMs), as traditional corpora provide limited information. Previous studies have synthesized and integrated corpora-dependent question-answering (QA) data to improve model performance but face challenges in QA data scalability and knowledge diversity, particularly in cross-domain contexts. Furthermore, leveraging our designed discipline and difficulty annotation system, we probe model deficiencies in STEM disciplines and high-difficulty data. To overcome these limitations, we propose a novel diversified pipeline to synthesize BoostQA, a 100B-token large-scale QA dataset. Our synthesis framework: (1) curates seed data from heterogeneous sources; (2) utilizes DeepSeek-R1 to implement STEM-focused multi-grade synthesis to boost data diversity and high-difficulty synthesis to mitigate difficulty degradation; (3) refines answers via DeepSeek-V3 to improve output quality. We utilize BoostQA in mid-training, a mid-stage between pre-training and post-training, to optimize domain-specific knowledge acquisition and enhance data quality. Our method enables Llama-3 8B, mid-trained on a 40B-token dataset, to achieve an average improvement of $\mathbf{12.74\%}$ on MMLU and CMMLU and establish SOTA average performance across 12 benchmarks. BoostQA also demonstrates robust scalability, with performance consistently improving as model size, data volume, and initial FLOPs scale.

Related papers

• Teaching Large Language Models to Maintain Contextual Faithfulness via Synthetic Tasks and Reinforcement Learning [80.27561080938747]
  We propose a systematic framework, CANOE, to improve the faithfulness of large language models (LLMs) in both short-form and long-form generation tasks without human annotations.<n>Also, we propose Dual-GRPO, a rule-based reinforcement learning method that includes three tailored rule-based rewards derived from synthesized short-form QA data.<n> Experimental results show that CANOE greatly improves the faithfulness of LLMs across 11 different downstream tasks, even outperforming the most advanced LLMs.
  arXiv  Detail & Related papers  (2025-05-22T10:10:07Z)
• Scaling Laws of Synthetic Data for Language Models [132.67350443447611]
  We introduce SynthLLM, a scalable framework that transforms pre-training corpora into diverse, high-quality synthetic datasets.<n>Our approach achieves this by automatically extracting and recombining high-level concepts across multiple documents using a graph algorithm.
  arXiv  Detail & Related papers  (2025-03-25T11:07:12Z)
• Will Pre-Training Ever End? A First Step Toward Next-Generation Foundation MLLMs via Self-Improving Systematic Cognition [89.50068130832635]
  Self-Improving cognition (SIcog) is a self-learning framework for constructing next-generation foundation MLLMs by multimodal knowledge.<n>We propose Chain-of-Description for step-by-step visual understanding and integrate structured Chain-of-Thought (CoT) reasoning to support in-depth multimodal reasoning.<n>Experiments demonstrate SIcog's effectiveness in developing MLLMs with enhanced multimodal cognition.
  arXiv  Detail & Related papers  (2025-03-16T00:25:13Z)
• Unleashing LLM Reasoning Capability via Scalable Question Synthesis from Scratch [54.12139707822201]
  We propose ScaleQuest, a novel, scalable, and cost-effective data synthesis method.<n>By generating diverse questions from scratch, we produce a dataset of 1 million problem-solution pairs.<n>Our experiments demonstrate that models trained on our data outperform existing open-source datasets.
  arXiv  Detail & Related papers  (2024-10-24T12:42:04Z)
• Data Quality Control in Federated Instruction-tuning of Large Language Models [43.29678396558287]
  Federated Learning enables privacy-preserving collaborative instruction tuning of large language models.<n>Local clients lack global visibility to filter noisy or low-quality samples before training.<n>We propose FedDQC, a novel federated instruction tuning framework with dynamic data quality control.
  arXiv  Detail & Related papers  (2024-10-15T12:14:57Z)
• Multi-OCT-SelfNet: Integrating Self-Supervised Learning with Multi-Source Data Fusion for Enhanced Multi-Class Retinal Disease Classification [2.5091334993691206]
  Development of a robust deep-learning model for retinal disease diagnosis requires a substantial dataset for training. The capacity to generalize effectively on smaller datasets remains a persistent challenge. We've combined a wide range of data sources to improve performance and generalization to new data.
  arXiv  Detail & Related papers  (2024-09-17T17:22:35Z)
• Semi-Supervised Reward Modeling via Iterative Self-Training [52.48668920483908]
  We propose Semi-Supervised Reward Modeling (SSRM), an approach that enhances RM training using unlabeled data. We demonstrate that SSRM significantly improves reward models without incurring additional labeling costs. Overall, SSRM substantially reduces the dependency on large volumes of human-annotated data, thereby decreasing the overall cost and time involved in training effective reward models.
  arXiv  Detail & Related papers  (2024-09-10T22:57:58Z)
• Towards Effective and Efficient Continual Pre-training of Large Language Models [163.34610964970258]
  Continual pre-training (CPT) has been an important approach for adapting language models to specific domains or tasks. This paper presents a technical report for continually pre-training Llama-3 (8B) It significantly enhances the Chinese language ability and scientific reasoning ability of the backbone model.
  arXiv  Detail & Related papers  (2024-07-26T13:55:21Z)
• Improved Techniques for the Conditional Generative Augmentation of Clinical Audio Data [36.45569352490318]
  We propose a conditional generative adversarial neural network-based augmentation method which is able to synthesize mel spectrograms from a learned data distribution. We show that our method outperforms all classical audio augmentation techniques and previously published generative methods in terms of generated sample quality. The proposed model advances the state-of-the-art in the augmentation of clinical audio data and improves the data bottleneck for the design of clinical acoustic sensing systems.
  arXiv  Detail & Related papers  (2022-11-05T10:58:04Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.