Breaking the Compression Barrier: Cross-Architecture Compression Boundary Learning via Reverse Regrowth

2026-08-16

AI Summary

This study addresses the abrupt performance degradation and indeterminate compression limits encountered in ultra-sparse model compression. We propose BRIDGE, a novel framework that reframes compression as constructive boundary search. By employing a hierarchical reverse regeneration strategy that integrates coarse-grained layer selection with fine-grained parameter recovery, BRIDGE precisely identifies and extends compression boundaries for both CNNs and Transformers. Experimental results demonstrate performance improvements of 1.49% and 4.77% under unstructured and structured pruning, respectively. These findings indicate that BRIDGE effectively surpasses existing compression limits while maintaining model stability, thereby establishing a new paradigm for extreme compression across diverse architectures.

Abstract

Model compression is critical for deploying networks on resource-constrained edge devices. While pruning-based methods can significantly reduce model size, they often suffer from abrupt performance collapse beyond a sparsity thresh-old, making it difficult to identify the feasible compression limit of the model. To address this challenge, we propose a boundary-Learning reverse regrowth framework, BRIDGE, that reformulates compression as a constructive boundary-search problem. Unlike forward pruning, our method first drives the model to an extremely sparse state to expose the collapse region, and then selectively regenerates the critical structure to restore performance. The proposed framework employs a hierarchical regeneration strategy, including coarse-grained layer selection and fine-grained regeneration parameter selection, to accurately identify which parameters require recovery. Experiments show that our method can recover models from the brink of collapse on both CNNs and Transformer architectures, demonstrating its architecture in-dependence. BRIDGE achieves a performance improvement of up to 1.49% in unstructured pruning and up to 4.77% in structured pruning. These results demonstrate that reverse regeneration can effectively extend the compression limit while maintaining stable performance. The source code is available at https://github.com/EnumaCaliber/BRIDGE.

Problem

• Model Compression
• Pruning
• Performance Collapse
• Compression Limit
• Sparsity Threshold

Innovation

• Reverse Regrowth
• Compression Boundary Learning
• Hierarchical Regeneration
• Cross-Architecture
• Model Compression