Cover Image

VisionTrim: Unified Vision Token Compression for Training-Free MLLM Acceleration

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Yu, Hanxun, Li, Wentong, Qu, Xuan, Wang, Song, Chen, Junbo, Zhu, Jianke
Format:	Preprint
Published:	2026
Subjects:	Computer Vision and Pattern Recognition
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

Similar Items

Inst3D-LMM: Instance-Aware 3D Scene Understanding with Multi-modal Instruction Tuning
by: Yu, Hanxun, et al.
Published: (2025)

Token Transforming: A Unified and Training-Free Token Compression Framework for Vision Transformer Acceleration
by: Zeng, Fanhu, et al.
Published: (2025)

Filter, Correlate, Compress: Training-Free Token Reduction for MLLM Acceleration
by: Han, Yuhang, et al.
Published: (2024)

Unlocking Dense Metric Depth Estimation in VLMs
by: Yu, Hanxun, et al.
Published: (2026)

Uncertainty-Instructed Structure Injection for Generalizable HD Map Construction
by: Liu, Xiaolu, et al.
Published: (2025)

MGMap: Mask-Guided Learning for Online Vectorized HD Map Construction
by: Liu, Xiaolu, et al.
Published: (2024)

ReliOcc: Towards Reliable Semantic Occupancy Prediction via Uncertainty Learning
by: Wang, Song, et al.
Published: (2024)

UniCompress: Token Compression for Unified Vision-Language Understanding and Generation
by: Wang, Ziyao, et al.
Published: (2026)

Not All Voxels Are Equal: Hardness-Aware Semantic Scene Completion with Self-Distillation
by: Wang, Song, et al.
Published: (2024)

Label-efficient Semantic Scene Completion with Scribble Annotations
by: Wang, Song, et al.
Published: (2024)

PVC: Progressive Visual Token Compression for Unified Image and Video Processing in Large Vision-Language Models
by: Yang, Chenyu, et al.
Published: (2024)

TokenPacker: Efficient Visual Projector for Multimodal LLM
by: Li, Wentong, et al.
Published: (2024)

EfficientVLA: Training-Free Acceleration and Compression for Vision-Language-Action Models
by: Yang, Yantai, et al.
Published: (2025)

A Coarse-to-Fine Approach to Multi-Modality 3D Occupancy Grounding
by: Shi, Zhan, et al.
Published: (2025)

StereoVGGT: A Training-Free Visual Geometry Transformer for Stereo Vision
by: Chen, Ziyang, et al.
Published: (2026)

PointLoRA: Low-Rank Adaptation with Token Selection for Point Cloud Learning
by: Wang, Song, et al.
Published: (2025)

HVOFusion: Incremental Mesh Reconstruction Using Hybrid Voxel Octree
by: Liu, Shaofan, et al.
Published: (2024)

AToken: A Unified Tokenizer for Vision
by: Lu, Jiasen, et al.
Published: (2025)

Unified Vision-Language-Action Model
by: Wang, Yuqi, et al.
Published: (2025)

Training-free Token Reduction for Vision Mamba
by: Ma, Qiankun, et al.
Published: (2025)

Fine-Grained Multi-View Hand Reconstruction Using Inverse Rendering
by: Gan, Qijun, et al.
Published: (2024)

RedVTP: Training-Free Acceleration of Diffusion Vision-Language Models Inference via Masked Token-Guided Visual Token Pruning
by: Xu, Jingqi, et al.
Published: (2025)

Forging Spatial Intelligence: A Roadmap of Multi-Modal Data Pre-Training for Autonomous Systems
by: Wang, Song, et al.
Published: (2025)

On the Limits of Token Reduction for Efficient Unified Vision Language Training
by: Chen, Siyi, et al.
Published: (2026)

VLA-InfoEntropy: A Training-Free Vision-Attention Information Entropy Approach for Vision-Language-Action Models Inference Acceleration and Success
by: Liu, Chuhang, et al.
Published: (2026)

MambaMap: Online Vectorized HD Map Construction using State Space Model
by: Yang, Ruizi, et al.
Published: (2025)

DynFlowDrive: Flow-Based Dynamic World Modeling for Autonomous Driving
by: Liu, Xiaolu, et al.
Published: (2026)

BackdoorVLM: A Benchmark for Backdoor Attacks on Vision-Language Models
by: Li, Juncheng, et al.
Published: (2025)

Multi-Token Enhancing for Vision Representation Learning
by: Li, Zhong-Yu, et al.
Published: (2024)

ST$^3$: Accelerating Multimodal Large Language Model by Spatial-Temporal Visual Token Trimming
by: Zhuang, Jiedong, et al.
Published: (2024)

VLA-IAP: Training-Free Visual Token Pruning via Interaction Alignment for Vision-Language-Action Models
by: Cheng, Jintao, et al.
Published: (2026)

VisionSelector: End-to-End Learnable Visual Token Compression for Efficient Multimodal LLMs
by: Zhu, Jiaying, et al.
Published: (2025)

Unified Language-Vision Pretraining in LLM with Dynamic Discrete Visual Tokenization
by: Jin, Yang, et al.
Published: (2023)

Vision to Geometry: 3D Spatial Memory for Sequential Embodied MLLM Reasoning and Exploration
by: Cai, Zhongyi, et al.
Published: (2025)

ILLUME+: Illuminating Unified MLLM with Dual Visual Tokenization and Diffusion Refinement
by: Huang, Runhui, et al.
Published: (2025)

HSD: Training-Free Acceleration for Document Parsing Vision-Language Model with Hierarchical Speculative Decoding
by: Liao, Wenhui, et al.
Published: (2026)

N3D-VLM: Native 3D Grounding Enables Accurate Spatial Reasoning in Vision-Language Models
by: Wang, Yuxin, et al.
Published: (2025)

MADTP: Multimodal Alignment-Guided Dynamic Token Pruning for Accelerating Vision-Language Transformer
by: Cao, Jianjian, et al.
Published: (2024)

HybridToken-VLM: Hybrid Token Compression for Vision-Language Models
by: Zhang, Jusheng, et al.
Published: (2025)

Top-Down Compression: Revisit Efficient Vision Token Projection for Visual Instruction Tuning
by: li, Bonan, et al.
Published: (2025)