ZIP-FIT: Embedding-Free Data Selection via Compression-Based Alignment

TL;DR

ZIP-FIT introduces a compression-based data selection method that effectively aligns training data with target tasks, leading to faster and more efficient language model training for specialized tasks like Autoformalization and code generation.

Contribution

The paper presents ZIP-FIT, a novel data selection framework using gzip compression to measure task alignment, outperforming existing methods in efficiency and effectiveness.

Findings

ZIP-FIT achieves up to 85.1% faster training convergence.

ZIP-FIT selection is up to 65.8% faster than DSIR.

Small, well-aligned datasets outperform larger, less targeted ones.

Abstract

Data selection is crucial for optimizing language model (LM) performance on specific tasks, yet most existing methods fail to effectively consider the target task distribution. Current approaches either ignore task-specific requirements entirely or rely on approximations that fail to capture the nuanced patterns needed for tasks like Autoformalization or code generation. Methods that do consider the target distribution often rely on simplistic, sometimes noisy, representations, like hashed n-gram features, which can lead to collisions and introduce noise. We introduce ZIP-FIT, a data selection framework that uses gzip compression to directly measure alignment between potential training data and the target task distribution. In extensive evaluations on Autoformalization and Python code generation, ZIP-FIT significantly outperforms leading baselines like DSIR and D4. Models trained on ZIP-FIT-selected data achieve their lowest cross-entropy loss up to 85.1\% faster than baselines, demonstrating that better task alignment leads to more efficient learning. In addition, ZIP-FIT performs selection up to 65.8\% faster than DSIR and two orders of magnitude faster than D4. Notably, ZIP-FIT shows that smaller, well-aligned datasets often outperform larger but less targeted ones, demonstrating that a small amount of higher quality data is superior to a large amount of lower quality data. Our results imply that task-aware data selection is crucial for efficient domain adaptation, and that compression offers a principled way to measure task alignment. By showing that targeted data selection can dramatically improve task-specific performance, our work provides new insights into the relationship between data quality, task alignment, and model learning efficiency.