Programming Every Example: Lifting Pre-training Data Quality Like Experts at Scale

TL;DR

ProX enables small language models to refine training data at scale by treating data cleaning as a programming task, leading to improved model performance and efficiency across diverse benchmarks and domains.

Contribution

The paper introduces ProX, a novel framework that allows models to generate and execute data refinement operations, outperforming traditional human-crafted rules and filtering methods.

Findings

Models trained on ProX-curated data outperform original and filtered data by over 2%.

ProX improves domain-specific pre-training accuracy by up to 20%.

ProX significantly reduces training FLOPs, enhancing efficiency.

Abstract

Large language model pre-training has traditionally relied on human experts to craft heuristics for improving the corpora quality, resulting in numerous rules developed to date. However, these rules lack the flexibility to address the unique characteristics of individual example effectively. Meanwhile, applying tailored rules to every example is impractical for human experts. In this paper, we demonstrate that even small language models, with as few as 0.3B parameters, can exhibit substantial data refining capabilities comparable to those of human experts. We introduce Programming Every Example (ProX), a novel framework that treats data refinement as a programming task, enabling models to refine corpora by generating and executing fine-grained operations, such as string normalization, for each individual example at scale. Experimental results show that models pre-trained on ProX-curated data outperform either original data or data filtered by other selection methods by more than 2% across various downstream benchmarks. Its effectiveness spans various model sizes and pre-training corpora, including C4, RedPajama-V2, FineWeb, FineWeb-Edu, and DCLM. Furthermore, ProX exhibits significant potential in domain-specific continual pre-training: without domain specific design, models trained on OpenWebMath refined by ProX outperform human-crafted rule-based methods, improving average accuracy by 7.6% over Mistral-7B, with 14.6% for Llama-2-7B and 20.3% for CodeLlama-7B, all within 10B tokens to be comparable to models like Llemma-7B trained on 200B tokens. Further analysis highlights that ProX significantly saves training FLOPs, offering a promising path for efficient LLM pre-training. We are open-sourcing ProX with >500B corpus, models, and sharing all training and implementation details for reproducible research and future innovation. Code: https://github.com/GAIR-NLP/ProX