AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

Feather the Throttle: Revisiting Visual Token Pruning for Vision-Language Model Acceleration

Mark Endo, Xiaohan Wang, Serena Yeung-Levy

ICCV 2025 · Vision-language and multimodal › Multimodal LLMs · approach: Attention-based visual token pruning · 20 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
remove RoPE when computing attention pruning criterion (ablation)FastV criterion · Localization avg (pruning after layer 3)accuracy: 5.9 -> 16.7Improved
remove RoPE when computing attention pruning criterion (ablation)FastV criterion · Localization avg (pruning after layer 8)accuracy: 23.3 -> 27.3Improved
remove RoPE when computing attention pruning criterion (ablation)FastV criterion · Open-ended VQA avg (pruning after layer 3)accuracy: 54.8 -> 59.0Improved
prune after layer 8 instead of layer 3 (ablation)phi_-R · Localization avg (final method comparison) (pruning layers 3 vs 8)accuracy: 16.7 -> 27.3Improved
add uniform sampling to RoPE-free attention criterion (ablation)phi_-R · Localization avg (pruning after layer 3)accuracy: 16.7 -> 27.2Improved
add uniform sampling to RoPE-free attention criterion (ablation)phi_-R · Localization avg (pruning after layer 8)accuracy: 27.3 -> 35.6Improved
combine phi_-R with uniform sampling versus uniform alone (ablation)phi_uniform · Localization avg (pruning after layer 3)accuracy: 28.0 -> 27.2Hurt
combine phi_-R with uniform sampling versus uniform alone (ablation)phi_uniform · Localization avg (pruning after layer 8)accuracy: 30.3 -> 35.6Improved
use tokens selected after layer 8 instead of layer 3 for whole LLM (ablation)phi_-R · OCID-Refaccuracy: 23.8 -> 26.7Improved
replace FastV with FeatHer at similar FLOPS reductionFastV · Localization avg (final method comparison)accuracy: 5.9 -> 39.3Improved
replace PyramidDrop with FeatHer at similar FLOPS reductionPyramidDrop · Localization avg (final method comparison)accuracy: 28.9 -> 39.3Improved
shuffle image token positions (ablation)unpruned baseline · Localization avgaccuracy: 53.2 -> 8.0Hurt

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