AI Paper Atlas

Research map / Architectures and efficiency / Pruning and efficient tuning

Large Language Model Compression with Global Rank and Sparsity Optimization

Changhai Zhou, Qian Qiao, Yuhua Zhou, Yuxin Wu, Shichao Weng, WEIZHONG ZHANG, Cheng Jin

ICLR 2026 · Architectures and efficiency › Pruning and efficient tuning · approach: Layer-wise pruning of LLMs · 8 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use CAP (RPCA + learned global allocation) instead of WandaLLaMA-3.1-8B-Instruct 50% sparse · GSM8Kaccuracy (%): 45.6 -> 56.8Improved
use CAP instead of OATS at 50% sparsityLLaMA-3 8B · WikiTextperplexity: 10.87 -> 10.35Improved
use CAP instead of SparseGPT at 50% sparsityPhi-3 Mini · zero-shot avgaccuracy (%): 66.36 -> 69.12Improved
use CAP instead of layerwise allocation (OWL) on WandaLLaMA-3 8B 50% · zero-shot avgaccuracy (%): 63.95 -> 66.85Improved
add sparse component to pure SVD compressionLLaMA-7B 20% · WikiText-2perplexity: 6.08 -> 5.85Improved
add fine-tuning to CAPDeBERTa-V3-base · MNLIaccuracy (%): 78.2 -> 86.8Improved
training-free CAP vs LoSparse (with FT)DeBERTa-V3-base · MNLIaccuracy (%): 83.8 -> 78.2Hurt
replace learned global allocation with uniform thresholding of RPCA output (ablation)CAP · LLaMA2-7BperformanceHurt
use CAP's high-sparsity sparse component for inferenceLLaMA-3.1-8B · throughputtok/s: 163.4 -> 176.5Improved
calibrate on WikiText-2 instead of C4 (ablation)CAP · LLaMA-3.1-8Bperplexity: PPL 7.05 -> 6.88; acc 76.8 -> 75.4Mixed
calibrate on GitHub Code instead of C4 (ablation)CAP · LLaMA-3.1-8Baccuracy (%): 76.8 -> 76.1No clear effect
use CAP on GLUE with BERT-base at 16% budget vs CoFiBERT-base · MNLIaccuracy (%): 80.5 -> 82.8Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy