Frequency Matters: Fast Model-Agnostic Data Curation for Pruning and Quantization

arXiv:2603.16105v2 Announce Type: replace-cross
Abstract: Post-training model compression is essential for enhancing the portability of Large Language Models (LLMs) while preserving their performance. While several compression approaches have been proposed, less emphasis has been placed on selecting the most suitable set of data (the so-called emphcalibration data) for finding the compressed model configuration. The choice of calibration data is a critical step in preserving model capabilities both intra- and inter-tasks. In this work, we address the challenge of identifying high-performance calibration sets for both pruning and quantization by analyzing intrinsic data properties rather than model-specific signals. We introduce texttttextbfZipCal, a model-agnostic data curation strategy that maximizes lexical diversity based on Zipfian power laws. Experiments demonstrate that our method consistently outperforms standard uniform random sampling across various pruning benchmarks. Notably, it also performs on par, in terms of downstream performance, with a state-of-the-art method that relies on model perplexity. The latter becomes prohibitively expensive at large-scale models and datasets, while texttttextbfZipCal is on average $sim$240$times$ faster due to its tractable linear complexityfootnoteWe make the code and the experiments available at https://github.com/FrancescoMonaco/ZipCal..