Let LLM Tell What to Prune and How Much to Prune

Mingzhe Yang, Sihao Lin, Changlin Li, Xiaojun Chang
Proceedings of the 42nd International Conference on Machine Learning, PMLR 267:70833-70849, 2025.

Abstract

Large language models (LLMs) have revolutionized various AI applications. However, their billions of parameters pose significant challenges for practical deployment. Structured pruning is a hardware-friendly compression technique and receives widespread attention. Nonetheless, existing literature typically targets a single structure of LLMs. We observe that the structure units of LLMs differ in terms of inference cost and functionality. Therefore, pruning a single structure unit in isolation often results in an imbalance between performance and efficiency. In addition, previous works mainly employ a prescribed pruning ratio. Since the significance of LLM modules may vary, it is ideal to distribute the pruning load to a specific structure unit according to its role within LLMs. To address the two issues, we propose a pruning method that targets multiple LLM modules with dynamic pruning ratios. Specifically, we find the intrinsic properties of LLMs can guide us to determine the importance of each module and thus distribute the pruning load on demand, i.e., what to prune and how much to prune. This is achieved by quantifying the complex interactions within LLMs. Extensive experiments on multiple benchmarks and LLM variants demonstrate that our method effectively balances the trade-off between efficiency and performance.

Cite this Paper

BibTeX
@InProceedings{pmlr-v267-yang25p, title = {Let {LLM} Tell What to Prune and How Much to Prune}, author = {Yang, Mingzhe and Lin, Sihao and Li, Changlin and Chang, Xiaojun}, booktitle = {Proceedings of the 42nd International Conference on Machine Learning}, pages = {70833--70849}, year = {2025}, editor = {Singh, Aarti and Fazel, Maryam and Hsu, Daniel and Lacoste-Julien, Simon and Berkenkamp, Felix and Maharaj, Tegan and Wagstaff, Kiri and Zhu, Jerry}, volume = {267}, series = {Proceedings of Machine Learning Research}, month = {13--19 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v267/main/assets/yang25p/yang25p.pdf}, url = {https://proceedings.mlr.press/v267/yang25p.html}, abstract = {Large language models (LLMs) have revolutionized various AI applications. However, their billions of parameters pose significant challenges for practical deployment. Structured pruning is a hardware-friendly compression technique and receives widespread attention. Nonetheless, existing literature typically targets a single structure of LLMs. We observe that the structure units of LLMs differ in terms of inference cost and functionality. Therefore, pruning a single structure unit in isolation often results in an imbalance between performance and efficiency. In addition, previous works mainly employ a prescribed pruning ratio. Since the significance of LLM modules may vary, it is ideal to distribute the pruning load to a specific structure unit according to its role within LLMs. To address the two issues, we propose a pruning method that targets multiple LLM modules with dynamic pruning ratios. Specifically, we find the intrinsic properties of LLMs can guide us to determine the importance of each module and thus distribute the pruning load on demand, i.e., what to prune and how much to prune. This is achieved by quantifying the complex interactions within LLMs. Extensive experiments on multiple benchmarks and LLM variants demonstrate that our method effectively balances the trade-off between efficiency and performance.} }
Endnote
%0 Conference Paper %T Let LLM Tell What to Prune and How Much to Prune %A Mingzhe Yang %A Sihao Lin %A Changlin Li %A Xiaojun Chang %B Proceedings of the 42nd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2025 %E Aarti Singh %E Maryam Fazel %E Daniel Hsu %E Simon Lacoste-Julien %E Felix Berkenkamp %E Tegan Maharaj %E Kiri Wagstaff %E Jerry Zhu %F pmlr-v267-yang25p %I PMLR %P 70833--70849 %U https://proceedings.mlr.press/v267/yang25p.html %V 267 %X Large language models (LLMs) have revolutionized various AI applications. However, their billions of parameters pose significant challenges for practical deployment. Structured pruning is a hardware-friendly compression technique and receives widespread attention. Nonetheless, existing literature typically targets a single structure of LLMs. We observe that the structure units of LLMs differ in terms of inference cost and functionality. Therefore, pruning a single structure unit in isolation often results in an imbalance between performance and efficiency. In addition, previous works mainly employ a prescribed pruning ratio. Since the significance of LLM modules may vary, it is ideal to distribute the pruning load to a specific structure unit according to its role within LLMs. To address the two issues, we propose a pruning method that targets multiple LLM modules with dynamic pruning ratios. Specifically, we find the intrinsic properties of LLMs can guide us to determine the importance of each module and thus distribute the pruning load on demand, i.e., what to prune and how much to prune. This is achieved by quantifying the complex interactions within LLMs. Extensive experiments on multiple benchmarks and LLM variants demonstrate that our method effectively balances the trade-off between efficiency and performance.
APA
Yang, M., Lin, S., Li, C. & Chang, X.. (2025). Let LLM Tell What to Prune and How Much to Prune. Proceedings of the 42nd International Conference on Machine Learning, in Proceedings of Machine Learning Research 267:70833-70849 Available from https://proceedings.mlr.press/v267/yang25p.html.

Related Material