AI Paper Atlas

Research map / Architectures and efficiency / Pruning and efficient tuning

GEM: A Scale-Aware and Distribution-Sensitive Sparse Fine-Tuning Framework for Effective Downstream Adaptation

Sungmin Kang, Jisoo Kim, Salman Avestimehr, Sunwoo Lee

AAAI 2026 · Architectures and efficiency › Pruning and efficient tuning · approach: Adapter-based parameter-efficient tuning · 1 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
add entropy to norm-based layer importance (ablation)GEM (OPT-1.3B) · MultiRCvalidation accuracy: 66.56% -> 70.31%Improved
add entropy to norm-based layer importance (ablation)GEM (OPT-125M) · SST-2accuracy: 87.12 -> 89.53Improved
use entropy-guided allocation instead of uniform per-layer allocation (ablation)GEM (Phi-2) · SQuADF1: 88.70 -> 90.89Improved
use norm only (without entropy) layer score instead of uniform (ablation)GEM (Phi-2) · SST-2accuracy: 93.37 -> 92.66Hurt
use entropy only (no norm) layer score (ablation)GEM (OPT-1.3B) · SST-2accuracy: 91.98 -> 91.92 (vs Uniform)No clear effect
select parameters by gradient-to-weight ratio instead of gradient magnitude (ablation)GEM (OPT-1.3B, 0.1%) · 7-task averageaverage score: 71.42 -> 77.47Improved
select parameters by gradient-to-weight ratio instead of randomGEM (OPT-125M, 0.1%) · 7-task averageaverage score: 66.29 -> 68.85Improved
reduce tunable ratio below 0.1% (ablation)GEM (Phi-2) · 7-task averageaverage scoreHurt
use GEM instead of full fine-tuningPhi-2, 0.1% params · GSM8kexact match: 53.41 -> 48.78Hurt
use GEM instead of full fine-tuningPhi-2, 0.1% params · MBPPpass@3: 47.83 -> 43.58Hurt
use GEM instead of full fine-tuningOPT-125M/1.3B/Phi-2, 0.1% params · 7-task averageaverage score: OPT-125M 67.23 -> 68.85; OPT-1.3B 76.96 -> 77.47; Phi-2 85.27 -> 85.52Mixed
use GEM instead of AdaLoRAPhi-2 · 7-task averageaverage score: 84.64 -> 85.52Improved

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