AI Paper Atlas

Research map / Architectures and efficiency / Pruning and efficient tuning

GHOST: Unmasking Phantom States in Mamba2 via Grouped Hidden-state Output-aware Selection & Truncation

Michael Menezes, Anastasios Kyrillidis

ICML 2026 · Architectures and efficiency › Pruning and efficient tuning · approach: Layer-wise pruning of LLMs · 0 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 joint controllability x observability saliency instead of controllability only (ablation)GHOST · WikiText-2 (50% sparsity)perplexity: 14.56 -> 14.23Improved
use joint saliency instead of observability only (ablation)GHOST · WikiText-2 (50% sparsity)perplexity: 16.16 -> 14.23Improved
use joint saliency instead of controllability only at 90% sparsity (ablation)GHOST · WikiText-2 (90% sparsity)perplexity: 2158 -> 25.07Improved
use joint saliency instead of controllability only at 70% sparsity (ablation)GHOST · WikiText-2 (70% sparsity)perplexity: 16.71 -> 16.16Improved
replace one-shot Taylor scoring with GHOST forward-pass sequential scoringMamba2-1.3B · WikiText-2 (70% sparsity)perplexity: 4255 -> 16.16Improved
use GHOST instead of Taylor on small modelsMamba2-130M · WikiText-2 (50% sparsity)perplexity: 1E16 -> 29.06Improved
use GHOST instead of Taylor on large modelsMamba2-2.7B · WikiText-2 (50% sparsity)perplexity: 11.50 -> 12.14Hurt
calibrate on short context (128) and evaluate on 2048 using GHOST vs TaylorMamba2-1.3B · WikiText-2 length generalizationperplexity: 1613 -> 14.27Improved
use GHOST instead of Magnitude pruningMamba2-1.3B · WikiText-2 (50% sparsity)perplexity: 29.34 -> 14.23Improved
increase number of calibration samples from 8 to 128 (ablation)GHOST · WikiText-2 (50% sparsity)perplexity: 14.28 -> 14.23No clear effect
prune Mamba2 state with GHOST instead of Taylor on LambadaMamba2-1.3B · Lambada zero-shotaccuracy: 59.23% (Taylor) vs 14-point drop from denseHurt
GHOST instead of Taylor on OOD mathMamba2-1.3B · MMLU Elementary Mathaccuracy: 22.49 -> 25.66Improved

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