Scaling Laws for Parameter Pruning in LLMs

ICLR 2026 Conference Submission16374 Authors

Keywords: Scaling laws, Model Compression, Large Language Models

Abstract: Scaling up model parameters and training data consistently improves the performance of large language models (LLMs), but at the cost of rapidly growing memory and compute requirements, which makes deployment on resource-limited hardware infeasible. *Model pruning*, a widely used compression technique, reduces inference costs by removing redundant parameters. However, its impact on downstream performance remains unpredictable and is typically assessed only through costly empirical sweeps. To address this gap, we introduce *pruning laws* -- simple and interpretable scaling relations that connect a pruned LLM's post-pruning performance to its unpruned performance and pruning ratio. Across five LLMs (2.7B–13B parameters), three pruning strategies (unstructured, width, and depth), and eight diverse tasks, we show that pruning laws achieve strong predictive accuracy (average extrapolation error $<7$%), reliably quantify performance degradation, and identify critical pruning thresholds beyond which recovery is infeasible. Moreover, we demonstrate that the same laws transfer universally across architectures, pruning methods, and even unseen models in zero-shot and one-shot setups. These results provide both researchers and practitioners with a principled framework to select pruning strategies, estimate safe pruning ratios without exhaustive tuning, and deploy LLMs efficiently under real-world compute and latency constraints.

Primary Area: foundation or frontier models, including LLMs

Submission Number: 16374