AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

Growing a Twig to Accelerate Large Vision-Language Models

Zhenwei Shao, Mingyang Wang, Zhou Yu, Wenwen Pan, Yan Yang, Tao Wei, Hongyuan Zhang, Ning Mao, Wei Chen, Jun Yu

ICCV 2025 · Vision-language and multimodal › Multimodal LLMs · approach: Attention-based visual token pruning · 6 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
guide token pruning with last twig layer attention instead of backbone layer K (ablation)TwigVLM · 6 VLM benchmarksRelAcc: 82.3 -> 96.0Improved
use twig layer attention instead of backbone layer K+T at same depth (ablation)TwigVLM · 6 VLM benchmarksRelAcc: 86.2 -> 96.0Improved
add SSD to token pruning (TTP & SSD vs FastV) (ablation)TwigVLM · MM-VetRelSpd: 104.3 -> 153.6Improved
add self-speculative decoding without token pruning (ablation)TwigVLM · MM-VetRelSpd: 104.3 -> 146.7Improved
add TTP on top of SSD (ablation)TwigVLM · MM-VetRelSpd: 146.7 -> 153.6Improved
initialize twig from layers K..K+T instead of random (ablation)TwigVLM · 6 VLM benchmarksRelAcc: 87.2 -> 96.0Improved
initialize twig from layers K..K+T instead of last T layers (ablation)TwigVLM · 6 VLM benchmarksRelAcc: 90.4 -> 96.0Improved
increase twig layers from 1 to 3 (ablation)TwigVLM · 6 VLM benchmarksRelAcc: 93.9 -> 96.0Improved
increase twig layers from 3 to 4 (ablation)TwigVLM · MM-Vet RelSpdRelSpd: 153.6 -> 145.4Hurt
prune at layer K=2 instead of K=0 (ablation)TwigVLM · 6 VLM benchmarksRelAcc: 93.6 -> 96.0Improved
prune at layer K=2 instead of K=3 (ablation)TwigVLM · 6 VLM benchmarksRelAcc: 85.5 -> 96.0Improved
add FinalWipe with Kf=24 (ablation)TwigVLM · 6 VLM benchmarksRelAcc: 93.1 -> 96.0Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 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