NIRVANA: Structured pruning reimagined for large language models compression

• URL: http://arxiv.org/abs/2509.14230v1
• Date: Wed, 17 Sep 2025 17:59:00 GMT
• Title: NIRVANA: Structured pruning reimagined for large language models compression
• Authors: Mengting Ai, Tianxin Wei, Sirui Chen, Jingrui He,
• Abstract summary: We introduce NIRVANA, a novel pruning method designed to balance immediate zero-shot preservation accuracy with robust fine-tuning.<n>To further address the unique challenges posed by structured pruning, NIRVANA incorporates an adaptive sparsity allocation mechanism across layers and modules.<n>Experiments conducted on Llama3, Qwen, T5 models demonstrate that NIRVANA outperforms existing structured pruning methods under equivalent sparsity constraints.
• Score: 50.651730342011014
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Structured pruning of large language models (LLMs) offers substantial efficiency improvements by removing entire hidden units, yet current approaches often suffer from significant performance degradation, particularly in zero-shot settings, and necessitate costly recovery techniques such as supervised fine-tuning (SFT) or adapter insertion. To address these critical shortcomings, we introduce NIRVANA, a novel pruning method explicitly designed to balance immediate zero-shot accuracy preservation with robust fine-tuning capability. Leveraging a first-order saliency criterion derived from the Neural Tangent Kernel under Adam optimization dynamics, NIRVANA provides a theoretically grounded pruning strategy that respects essential model training behaviors. To further address the unique challenges posed by structured pruning, NIRVANA incorporates an adaptive sparsity allocation mechanism across layers and modules (attention vs. MLP), which adjusts pruning intensity between modules in a globally balanced manner. Additionally, to mitigate the high sensitivity of pruning decisions to calibration data quality, we propose a simple yet effective KL divergence-based calibration data selection strategy, ensuring more reliable and task-agnostic pruning outcomes. Comprehensive experiments conducted on Llama3, Qwen, and T5 models demonstrate that NIRVANA outperforms existing structured pruning methods under equivalent sparsity constraints, providing a theoretically sound and practical approach to LLM compression. The code is available at https://github.com/iDEA-iSAIL-Lab-UIUC/NIRVANA.

Related papers

• Accuracy-Preserving CNN Pruning Method under Limited Data Availability [7.647276696906605]
  Convolutional Neural Networks (CNNs) are widely used in image recognition and have succeeded in various domains.<n>Research has been conducted on compressing pre-trained models for specific target applications in environments with limited computing resources.<n>This study proposes a pruning method that achieves a higher pruning rate while preserving better model accuracy.
  arXiv  Detail & Related papers  (2025-11-13T23:52:57Z)
• Elastic ViTs from Pretrained Models without Retraining [74.5386166956142]
  Vision foundation models achieve remarkable performance but are only available in a limited set of pre-determined sizes.<n>We introduce SnapViT: Single-shot network approximation for pruned Vision Transformers.<n>Our approach efficiently combines gradient information with cross-network structure correlations, approximated via an evolutionary algorithm.
  arXiv  Detail & Related papers  (2025-10-20T16:15:03Z)
• Adaptive Nonlinear Vector Autoregression: Robust Forecasting for Noisy Chaotic Time Series [0.0]
  vector autoregression and reservoir computing have shown promise in forecasting chaotic dynamical systems.<n>We propose an adaptive N model that combines delay-embedded linear inputs with features generated by a shallow, learnable multi-layer perceptron.
  arXiv  Detail & Related papers  (2025-07-11T16:40:10Z)
• Sample-aware Adaptive Structured Pruning for Large Language Models [14.605017410864583]
  This study introduces AdaPruner, a sample-aware adaptive structured pruning framework for large language models (LLMs)<n>Specifically, AdaPruner effectively removes redundant parameters from LLMs by constructing a structured pruning solution space.<n>At a 20% pruning ratio, the model pruned with AdaPruner maintains 97% of the performance of the unpruned model.
  arXiv  Detail & Related papers  (2025-03-08T12:00:21Z)
• Enhancing Reliability of Neural Networks at the Edge: Inverted Normalization with Stochastic Affine Transformations [0.22499166814992438]
  We propose a method to inherently enhance the robustness and inference accuracy of BayNNs deployed in in-memory computing architectures. Empirical results show a graceful degradation in inference accuracy, with an improvement of up to $58.11%$.
  arXiv  Detail & Related papers  (2024-01-23T00:27:31Z)
• Achieving Constraints in Neural Networks: A Stochastic Augmented Lagrangian Approach [49.1574468325115]
  Regularizing Deep Neural Networks (DNNs) is essential for improving generalizability and preventing overfitting. We propose a novel approach to DNN regularization by framing the training process as a constrained optimization problem. We employ the Augmented Lagrangian (SAL) method to achieve a more flexible and efficient regularization mechanism.
  arXiv  Detail & Related papers  (2023-10-25T13:55:35Z)
• Soft ascent-descent as a stable and flexible alternative to flooding [6.527016551650139]
  We propose a softened, pointwise mechanism called SoftAD that downweights points on the borderline, limits the effects of outliers, and retains the ascent-descent effect of flooding. We demonstrate how SoftAD can realize classification accuracy competitive with flooding while enjoying a much smaller loss generalization gap and model norm.
  arXiv  Detail & Related papers  (2023-10-16T02:02:56Z)
• Benign Overfitting in Deep Neural Networks under Lazy Training [72.28294823115502]
  We show that when the data distribution is well-separated, DNNs can achieve Bayes-optimal test error for classification. Our results indicate that interpolating with smoother functions leads to better generalization.
  arXiv  Detail & Related papers  (2023-05-30T19:37:44Z)
• GDP: Stabilized Neural Network Pruning via Gates with Differentiable Polarization [84.57695474130273]
  Gate-based or importance-based pruning methods aim to remove channels whose importance is smallest. GDP can be plugged before convolutional layers without bells and whistles, to control the on-and-off of each channel. Experiments conducted over CIFAR-10 and ImageNet datasets show that the proposed GDP achieves the state-of-the-art performance.
  arXiv  Detail & Related papers  (2021-09-06T03:17:10Z)
• Efficient Micro-Structured Weight Unification and Pruning for Neural Network Compression [56.83861738731913]
  Deep Neural Network (DNN) models are essential for practical applications, especially for resource limited devices. Previous unstructured or structured weight pruning methods can hardly truly accelerate inference. We propose a generalized weight unification framework at a hardware compatible micro-structured level to achieve high amount of compression and acceleration.
  arXiv  Detail & Related papers  (2021-06-15T17:22:59Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.