Cover Image

Multi-Granular Spatio-Temporal Token Merging for Training-Free Acceleration of Video LLMs

Fuente: arXiv

Saved in:

Bibliographic Details
Main Authors:	Hyun, Jeongseok, Hwang, Sukjun, Han, Su Ho, Kim, Taeoh, Lee, Inwoong, Wee, Dongyoon, Lee, Joon-Young, Kim, Seon Joo, Shim, Minho
Format:	Preprint
Published:	2025
Subjects:	Computer Vision and Pattern Recognition Artificial Intelligence
Online Access:	Acceder al recurso
Tags:	Add Tag No Tags, Be the first to tag this record!

Similar Items

Decomposed Attention Fusion in MLLMs for Training-Free Video Reasoning Segmentation
by: Han, Su Ho, et al.
Published: (2025)

Why Can't I Open My Drawer? Mitigating Object-Driven Shortcuts in Zero-Shot Compositional Action Recognition
by: Ahn, Geo, et al.
Published: (2026)

Classification Matters: Improving Video Action Detection with Class-Specific Attention
by: Lee, Jinsung, et al.
Published: (2024)

Video-Oasis: Rethinking Evaluation of Video Understanding
by: Lim, Geuntaek, et al.
Published: (2026)

Prototypes are Balanced Units for Efficient and Effective Partially Relevant Video Retrieval
by: Moon, WonJun, et al.
Published: (2025)

Exploring Scalability of Self-Training for Open-Vocabulary Temporal Action Localization
by: Hyun, Jeongseok, et al.
Published: (2024)

CoMoGaussian: Continuous Motion-Aware Gaussian Splatting from Motion-Blurred Images
by: Lee, Jungho, et al.
Published: (2025)

CoCoGaussian: Leveraging Circle of Confusion for Gaussian Splatting from Defocused Images
by: Lee, Jungho, et al.
Published: (2024)

ORBIS: Output-Guided Token Reduction with Distribution-Aware Matching for Video Diffusion Acceleration
by: Lee, Hangyeol, et al.
Published: (2026)

ActionSwitch: Class-agnostic Detection of Simultaneous Actions in Streaming Videos
by: Kang, Hyolim, et al.
Published: (2024)

A Simple Baseline with Single-encoder for Referring Image Segmentation
by: Yu, Seonghoon, et al.
Published: (2024)

VISAGE: Video Instance Segmentation with Appearance-Guided Enhancement
by: Kim, Hanjung, et al.
Published: (2023)

PHUMA: Physically-Grounded Humanoid Locomotion Dataset
by: Lee, Kyungmin, et al.
Published: (2025)

Learning to Enhance Aperture Phasor Field for Non-Line-of-Sight Imaging
by: Cho, In, et al.
Published: (2024)

Rethinking Token Reduction for Diffusion Models via Output-Similarity-Awareness
by: Lee, Hangyeol, et al.
Published: (2026)

Video Token Merging for Long-form Video Understanding
by: Lee, Seon-Ho, et al.
Published: (2024)

DiSC: Resolution-Scalable Acceleration of Diffusion Models by Exploiting Sparsity and Cached Token Reuse with Hash-based Distribution
by: Yoon, Jieon, et al.
Published: (2026)

Enhancing Topological Dependencies in Spatio-Temporal Graphs with Cycle Message Passing Blocks
by: Lee, Minho, et al.
Published: (2024)

NegMerge: Sign-Consensual Weight Merging for Machine Unlearning
by: Kim, Hyo Seo, et al.
Published: (2024)

SeiT++: Masked Token Modeling Improves Storage-efficient Training
by: Lee, Minhyun, et al.
Published: (2023)

Towards Scalable Handwriting Communication via EEG Decoding and Latent Embedding Integration
by: Kim, Jun-Young, et al.
Published: (2024)

Motion-Oriented Compositional Neural Radiance Fields for Monocular Dynamic Human Modeling
by: Kim, Jaehyeok, et al.
Published: (2024)

Enhancing Spatio-Temporal Zero-shot Action Recognition with Language-driven Description Attributes
by: Kim, Yehna, et al.
Published: (2025)

Domain Reduction Strategy for Non Line of Sight Imaging
by: Shim, Hyunbo, et al.
Published: (2023)

Coherent Human-Scene Reconstruction from Multi-Person Multi-View Video in a Single Pass
by: Kim, Sangmin, et al.
Published: (2026)

STELLA: Continual Audio-Video Pre-training with Spatio-Temporal Localized Alignment
by: Lee, Jaewoo, et al.
Published: (2023)

VideoMamba: Spatio-Temporal Selective State Space Model
by: Park, Jinyoung, et al.
Published: (2024)

Discrete Dictionary-based Decomposition Layer for Structured Representation Learning
by: Park, Taewon, et al.
Published: (2024)

ES-Merging: Biological MLLM Merging via Embedding Space Signals
by: Lee, Wonbin, et al.
Published: (2026)

Autoregressive Universal Video Segmentation Model
by: Heo, Miran, et al.
Published: (2025)

MASQ: Accelerating Masked Diffusion via Stage-Wise Multi-Precision Quantization
by: Kim, Seeyeon, et al.
Published: (2026)

Investigating Pre-Training Objectives for Generalization in Vision-Based Reinforcement Learning
by: Kim, Donghu, et al.
Published: (2024)

Adapting Pretrained ViTs with Convolution Injector for Visuo-Motor Control
by: Hwang, Dongyoon, et al.
Published: (2024)

Transport Dynamics of Water Molecules Confined between Lipid Membranes
by: Lee, Minho, et al.
Published: (2024)

Do's and Don'ts: Learning Desirable Skills with Instruction Videos
by: Kim, Hyunseung, et al.
Published: (2024)

SyMerge: From Non-Interference to Synergistic Merging via Single-Layer Adaptation
by: Jung, Aecheon, et al.
Published: (2024)

Optimized Layerwise Approximation for Efficient Private Inference on Fully Homomorphic Encryption
by: Lee, Junghyun, et al.
Published: (2023)

Local Representative Token Guided Merging for Text-to-Image Generation
by: Lee, Min-Jeong, et al.
Published: (2025)

Humans as a Calibration Pattern: Dynamic 3D Scene Reconstruction from Unsynchronized and Uncalibrated Videos
by: Choi, Changwoon, et al.
Published: (2024)

AutiHero: Engaging Parents in Creating Personalized, Multi-path Social Narratives for Autistic Children
by: Lee, Jungeun, et al.
Published: (2025)