UltraX: Refining Pre-Training Data at Scale with Adaptive Programmatic Editing

2026-07-09

AI Summary

This work addresses the challenges in large-scale pretraining data quality optimization, where rule-based methods lack flexibility and large-model-based approaches suffer from low efficiency and poor reliability. The authors propose UltraX, a novel framework that, for the first time, constructs a complete space of edit functions encompassing insertion, deletion, and modification operations to enable end-to-end programmable data refinement. By integrating expert large model–guided adaptive prompt optimization, structured program-supervised generation, line-aligned mapping, and dynamic context replacement, UltraX significantly enhances fine-grained controllability and scalability. Experiments demonstrate that UltraX achieves state-of-the-art performance across multiple corpora, matching or surpassing existing baselines with substantially fewer training tokens, thereby exhibiting exceptional data efficiency and refinement reliability.

Abstract

As available training data approaches its physical limit, gains from Scaling Laws have begun to diminish. Consequently, improving Large Language Models (LLMs) now depends less on data expansion and more on higher-quality data utilization. However, in the context of large-scale corpora, existing refinement methodologies face significant limitations in quality, efficiency, and reliability: Rule-based approaches are constrained by fixed heuristics and struggle with instance-level variations; LLM-based approaches improve quality but fail to meet the efficiency and reliability requirements of large-scale data processing. To address these challenges, we propose UltraX, a function-calling refinement framework for large-scale pre-training data that completes the editing function space by introducing insertion in addition to deletion and modification, enabling fine-grained instance-level editing. Specifically, UltraX builds a reliable program-supervision generation pipeline. In this pipeline, dataset-adaptive prompt optimization first guides an expert LLM to produce high-quality end-to-end refined texts, and Line Alignment Mapping and Dynamic Context Replacement then convert original-refined text pairs into structured program supervision. Meanwhile, UltraX improves supervision quality and stabilizes the training distribution with low-confidence example filtering and ratio-controlled sampling by operation combination. During inference and execution, it normalizes and validates model outputs through sliding-window prediction, global operation aggregation, and systematic post-processing, improving the stability and reliability of large-scale execution. Experiments show that UltraX achieves the highest average performance across all corpora and also matches or surpasses baselines with fewer training tokens, demonstrating stronger data efficiency and refinement reliability.

Problem

• pre-training data refinement
• large-scale corpora
• data quality
• efficiency
• reliability

Innovation

• adaptive programmatic editing
• fine-grained data refinement
• program supervision
• large-scale pre-training
• data efficiency