Challenge: Existing structured pruning methods fail to identify outlier-triggering tokens and uniform layer-wise sparsity misaligns with heterogeneous outlier distributions.
Approach: They propose a framework that prioritizes capturing outlier-triggering tokens rather than reconstructing full hidden distributions.
Outcome: Experiments on LLaMA2, LLama3 and OPT show that the proposed framework outperforms state-of-the-art methods and achieves 25% perplexity reduction at 1.6 speedup.

Similar Papers

CFSP: An Efficient Structured Pruning Framework for LLMs with Coarse-to-Fine Activation Information (2025.coling-main)

Copied to clipboard

Challenge: Existing LLM pruning works focus on unstructured pruning, which typically requires special hardware support for a practical speed-up.
Approach: They propose a network pruning framework that leverages both coarse and fine-grained activation information as an importance criterion to guide pruning.
Outcome: The proposed framework outperforms existing pruning methods on diverse models across sparsity budgets.
From Local to Global: Revisiting Structured Pruning Paradigms for Large Language Models (2026.acl-long)

Copied to clipboard

Challenge: Structured pruning is a practical approach to deploying large language models (LLMs) but it fails to capitalize on modest task-specific calibration signals, causing limited downstream gains.
Approach: They propose a method that removes attention heads and MLP channels using loss-based important scores . they use perplexity for language modeling and a margin-based objective for decision-style tasks .
Outcome: The proposed method lowers perplexity and improves accuracy at higher sparsity . it also stabilizes accuracy and mitigates perxity collapse without fine-tuning .
Pruning Weights but Not Truth: Safeguarding Truthfulness While Pruning LLMs (2025.findings-emnlp)

Copied to clipboard

Challenge: Neural network pruning disrupts LLMs’ internal activation features crucial for lie detection . layer-wise pruning sparsity inadvertently removes crucial weights, failing to improve lie detection performance despite its reliance on the most crucial LLM layer.
Approach: They propose a pruning approach that places greater emphasis on layers with more activation outliers and stronger discriminative features simultaneously.
Outcome: The proposed approach improves the hallucination detection for pruned LLMs (achieving 88% accuracy at 50% sparsity) and enhances their performance on TruthfulQA.
IG-Pruning: Input-Guided Block Pruning for Large Language Models (2025.emnlp-main)

Copied to clipboard

Challenge: Existing methods for efficient inference rely on fixed block masks, which can lead to suboptimal performance.
Approach: They propose an input-aware block-wise pruning method that dynamically selects layer masks at inference time.
Outcome: The proposed method outperforms state-of-the-art static depth pruning methods . it is particularly suitable for resource-constrained deployment scenarios .
Task-aware Block Pruning with Output Distribution Signals for Large Language Models (2026.findings-eacl)

Copied to clipboard

Challenge: Existing methods to estimate block importance rely on representation similarity or computationally expensive sensitivity analyses to estimate task-aware model behavior.
Approach: They propose a novel approach that quantifies block-level uncertainty from the statistics of each block’s early-exited output distribution on a calibration dataset.
Outcome: Experiments show that the proposed approach preserves downstream task performance while reducing inference latency and computational cost.
DRPruning: Efficient Large Language Model Pruning through Distributionally Robust Optimization (2025.acl-long)

Copied to clipboard

Challenge: Structured pruning reduces model size but often causes uneven degradation across domains, leading to biased performance.
Approach: They propose a method that dynamically adjusts the data distribution during training to restore balanced performance across heterogeneous and multi-tasking data.
Outcome: Experiments in monolingual and multilingual settings show that the proposed method surpasses similarly sized models in pruning and continued pretraining over perplexity, downstream tasks, and instruction tuning.
Structured Pruning for Diverse Best-of-N Reasoning Optimization (2025.findings-acl)

Copied to clipboard

Challenge: Extensive experiments on the MATH dataset demonstrate that our method significantly outperforms traditional best-of-N and random head selection strategies.
Approach: They propose a contrastive learning framework that dynamically selects the optimal head and layer to prune during inference by aligning question embeddings with head embedds.
Outcome: The proposed approach outperforms best-of-N and random head selection strategies on the MATH500 and GSM8K datasets.
GAP: a Global Adaptive Pruning Method for Large Language Models (2025.emnlp-main)

Copied to clipboard

Challenge: Existing structured pruning methods employ uniform compression rates across network layers, neglecting the varying importance of different network depths.
Approach: They propose a pruning framework that minimizes global capability loss by layer-adaptive pruning rates.
Outcome: The proposed approach achieves comparable performance with state-of-the-art methods at high pruning rates and shows significant advantages at low pruning rates.
Two-Stage Regularization-Based Structured Pruning for LLMs (2026.acl-long)

Copied to clipboard

Challenge: Structural pruning is a promising solution for large language models . prior structured pruning methods remove unimportant parameters based on certain metrics .
Approach: They propose a structural pruning method that iteratively learns the weights of transformer layers by adding their l1-norm to the loss function.
Outcome: The proposed pruning method outperforms strong layer-wise pruning methods without requiring retraining.
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.

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