Challenge: Existing structured pruning approaches for large language models require calibration data and costly continued pretraining on billions of tokens to recover lost performance.
Approach: They propose a method that locally distills activations with low-rank weights . they compress Mixtral-8x7B on a single GPU and Phi-2 3B by 40% .
Outcome: The proposed method compresses Mixtral-8x7B on a single A100 GPU, removing 10 billion parameters while retaining over 95% of its original performance.

Similar Papers

Pruning Large Language Models to Intra-module Low-rank Architecture with Transitional Activations (2024.findings-acl)

Copied to clipboard

Challenge: Structured pruning is a feasible solution for end-side LLM deployment . however, achieving a high compression ratio for scaled-up LLMs remains a challenge .
Approach: They propose a task-agnostic structured pruning approach coupled with a compact Transformer architecture to prune LLMs into an intra-module low-rank architecture.
Outcome: The proposed approach reduces transitional activations inside multi-head attention (MHA) and multi-layer perceptron (MLP) modules while preserving inter-module activations sensitive to perturbations.
Train Flat, Then Compress: Sharpness-Aware Minimization Learns More Compressible Models (2022.findings-emnlp)

Copied to clipboard

Challenge: Recent advances in hardware, modeling, and optimization for deep neural networks have led to improvements in memory and inference efficiency.
Approach: They propose to combine sharpness-aware minimization with various model compression methods to improve model compressibility.
Outcome: Empirically, optimizing for flatter minima leads to greater compressibility of parameters compared to vanilla Adam when fine-tuning BERT models, with little to no loss in accuracy on the GLUE text classification and SQuAD question answering benchmarks.
Focus on the Core: Efficient Attention via Pruned Token Compression for Document Classification (2023.findings-emnlp)

Copied to clipboard

Challenge: Pre-trained transformers suffer from a computationally expensive self-attention mechanism that interacts with all tokens, including those unfavorable to classification performance.
Approach: They propose to integrate token pruning and token combining strategies to improve model performance and reduce computational demands.
Outcome: Experiments with various datasets show that the proposed model performs better than baseline models, with the best improvement over the existing model.
Enhancing One-Shot Pruned Pre-trained Language Models through Sparse-Dense-Sparse Mechanism (2025.coling-main)

Copied to clipboard

Challenge: Pre-trained language models (PLMs) are robust in contextual understanding but their considerable size incurs significant computational and storage costs.
Approach: They propose a Sparse-Dense-Sparse pruning framework to prune PLMs . they prune less critical connections using conventional pruning methods .
Outcome: The proposed pruning framework outperforms SparseGPT and Wanda under identical sparsity.
Pruning Foundation Models for High Accuracy without Retraining (2024.findings-emnlp)

Copied to clipboard

Challenge: Despite the superior performance of foundation models, it is challenging to deploy large language models in practical applications due to their massive parameters and computations.
Approach: They propose a pruning algorithm to prune LLMs in one-shot without retraining . they propose retrainable pruning algorithms to prune multiple weights in LLM .
Outcome: The proposed pruning methods perform better than baseline pruning methods on sparse and unstructured sparsity models.
Structured Pruning for Efficient Generative Pre-trained Language Models (2023.findings-acl)

Copied to clipboard

Challenge: Large-scale generative Pre-trained Language Models (PLMs) are limited in their deployment in real-world applications.
Approach: They propose to prune the feed-forward networks of generative pre-trained language models to smaller widths without designing extra operators.
Outcome: The proposed method achieves 1.51x/6.96x inference speedup on GPU/CPU with 67% size reduction.
BWLA: Breaking the Barrier of W1AX Post-Training Quantization for LLMs (2026.acl-long)

Copied to clipboard

Challenge: Large language models have driven major progress in NLP, but memory and compute requirements hinder practical deployment.
Approach: They propose a framework that preserves high accuracy while achieving 1-bit weight quantization . the orthogonal-kronecker transformation learns an orthogonale mapping via EM minimization - a new approach to quantization is proposed .
Outcome: The proposed framework achieves 1-bit weight quantization with low activations with low-bit activations.
FLAT-LLM: Fine-grained Low-rank Activation Space Transformation for Large Language Model Compression (2026.findings-eacl)

Copied to clipboard

Challenge: Low-rank decomposition methods suffer from accuracy degradation and expensive calibration procedures.
Approach: They propose a fast and accurate, training-free structural compression method based on fine-grained low-rank transformations in the activation space.
Outcome: The proposed method outperforms pruning baselines in generalization and downstream performance while delivering inference speedups.
PRILoRA: Pruned and Rank-Increasing Low-Rank Adaptation (2024.findings-eacl)

Copied to clipboard

Challenge: Several approaches to parameter-efficient fine-tuning have been proposed . low-rank Adaptation (LoRA) does not consider the varying importance of each layer .
Approach: They propose a method that allocates a different rank for each layer and performs pruning throughout the training process.
Outcome: The proposed method is based on eight GLUE benchmarks and is currently the state of the art.
LoRAPrune: Structured Pruning Meets Low-Rank Parameter-Efficient Fine-Tuning (2024.findings-acl)

Copied to clipboard

Challenge: Low-rank adaption (LoRA) is a low-level pruning method that can be expensive and slow to deploy.
Approach: They propose a low-rank adaption pruning framework that provides an accurate structured pruned model in a memory-efficient manner.
Outcome: The proposed pruning framework reduces perplexity and memory usage by 52.6% on LLaMA and T5 models while reducing memory usage.

What is GenGO?

GenGO is an NLP powered publication search system. It currenctly indexes 30k+ papers from ACL Anthology, and implements multi-aspect summarization, semantic search, and more!

Information

About
Limitations