AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

FOLDER: Accelerating Multi-Modal Large Language Models with Enhanced Performance

Haicheng Wang, Zhemeng Yu, Gabriele Spadaro, Chen Ju, Victor Quétu, Shuai Xiao, Enzo Tartaglione

ICCV 2025 · Vision-language and multimodal › Multimodal LLMs · approach: Training-free token reduction for MLLMs · 30 citations (OpenAlex)

Read the paper

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
merge tokens only in the last vision block instead of uniformly in every block (ablation)FOLDER on BLIP · BLIP captioning (60% reduction)CIDEr: 128.8 -> 132.2Improved
merge tokens only in the last vision block instead of uniformly in every block (ablation)FOLDER on BLIP · BLIP captioning (70% reduction)CIDEr: 127.2 -> 131.0Improved
merge similar tokens by averaging instead of dropping them (ablation)FOLDER on BLIP · BLIP captioning (50% reduction)CIDEr: 130.0 -> 132.0Improved
use norm-weighted average instead of plain average (ablation)FOLDER on BLIP · BLIP captioning (50% reduction)CIDEr: 132.0 -> 132.2No clear effect
train LLaVA1.5-13B with FOLDER at 66% reductionLLaVA1.5-13B · 9-benchmark average, inference at 66%average score: 58.8 -> 61.2Improved
train LLaVA1.5-13B with FOLDER at 66% reductionLLaVA1.5-13B · Hallusion Benchaccuracy: 36.1 -> 43.6Improved
raise training and inference reduction from 66% to 75%LLaVA1.5-13B with FOLDER · 9-benchmark averageaverage score: 61.2 -> 58.2Hurt
use FOLDER instead of Turbo uniform merging at 66% reductionLLaVA1.5-7B · OCRVQAaccuracy: 34.4 -> 46.1Improved
use FOLDER instead of FastV at 66% reduction (average over benchmarks)LLaVA1.5-7B · 10-benchmark averageaverage score: 54.6 -> 54.8Mixed
use FOLDER instead of Turbo on Video-LLaVA at 65% reductionVideo-LLaVA · MME-Videoscore: 29.8 -> 31.9Improved
add FOLDER with 66% reduction to LLaVA1.5-13B trainingLLaVA1.5-13B · training timeGPU hours: 187 -> 124Improved
add FOLDER with 66% reduction before the LLMLLaVA1.5-13B · inference peak memorymemory: 81.4G -> 48.7GImproved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 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