Phase transitions in large language model compression
Abstract
This perspective argues that Large Language Models exhibit Model Phase Transitions: performance collapses beyond critical compression thresholds. We analyze structural, numerical, and algebraic redundancy across pruning, quantization, and low-rank decomposition techniques. These sources are orthogonal, enabling a criticality-aware compression framework that achieves near-lossless compression to 10% of the original size. This shift proves that compressing a giant is more effective than training a dwarf for efficient, sustainable AI deployment.
Fields of Study & Subjects
Topic Modeling
Multimodal Machine Learning Applications
Machine Learning in Materials Science
Publication details
- Journal / source
- npj Artificial Intelligence
- Publisher
- Springer Science and Business Media LLC
- Volume
- 2
- Issue
- 1
- Publication date
- 6 February 2026
- Language
- EN