AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

Scaling Capability in Token Space: An Analysis of Large Vision Language Model

Tenghui Li, Guoxu Zhou, Xuyang Zhao, Qibin Zhao

JMLR 2025 · Vision-language and multimodal › Multimodal LLMs · approach: Attention-based visual token pruning · 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
increase the number of learnable queries n_l (vision tokens) (ablation)Vision Question Queries model · OCRBenchfitted scaling exponent alpha: alpha -0.3727Improved
increase n_l (ablation)Vision Queries (ft) model · COCO VAL CIDErfitted scaling exponent alpha: alpha -0.1642 -> -0.3937Improved
increase n_l (ablation)Vision Question Queries model · RealWorldQAfitted scaling exponent alpha: alpha -0.0030No clear effect
increase n_l (ablation)Vision Queries (ft) model · RealWorldQAfitted scaling exponent alpha: alpha 0.0335No clear effect
increase n_l (ablation)Vision Question Queries model · MMStarfitted scaling exponent alpha: alpha -0.0272Improved
increase n_l (ablation)Vision Question Queries model · MMEfitted scaling exponent alpha: alpha -0.0516Improved
increase n_l (ablation)Vision Question Queries model · COCO VAL BLEU-4fitted scaling exponent alpha: alpha -0.1161Improved
remove question text from fusion input (ft vision queries only) (ablation)Vision Queries (ft) model · OCRBenchfitted scaling exponent alpha: alpha -0.3727 -> -0.4142Improved
remove question text from fusion input (ft vision queries only) (ablation)Vision Queries (ft) model · POPEfitted scaling exponent alpha: alpha -0.0503 -> -0.0377Hurt
remove question text from fusion input (ft vision queries only) (ablation)Vision Queries (ft) model · AI2D TESTfitted scaling exponent alpha: alpha -0.0359 -> -0.0902Improved
remove question text from fusion input (ft vision queries only) (ablation)Vision Queries (ft) model · SEEDBench IMGfitted scaling exponent alpha: alpha -0.0710 -> -0.0516Hurt

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