arXiv:2611.25574v1 cs.CV

VidToC: Rethinking Video Token Compression for Multimodal Large Language Models

Zihan Xiao (AIM3 Lab, Renmin University of China), Boshen Xu (AIM3 Lab, Renmin University of China), Pei Fu (MiLM Plus, Xiaomi Inc), Jianzhong Ju (MiLM Plus, Xiaomi Inc), Zhenbo Luo (MiLM Plus, Xiaomi Inc), Jian Luan (MiLM Plus, Xiaomi Inc), Qin Jin (AIM3 Lab, Renmin University of China)

제출 2026-11-20

원문 초록

Extending multimodal large language models (MLLMs) from images to videos faces substantial memory and latency challenges due to the massive number of visual tokens. These challenges have driven the development of numerous visual token compression methods and more efficient LLM architectures. However, existing compression methods are often evaluated with different training setups, tasks, and models, making fair comparison difficult. Meanwhile, the growing use of hybrid Mamba-Transformer architectures raises questions about whether compression strategies transfer across backbones. We present VidToC, a controlled evaluation framework for representative compression algorithms, LLM backbones, and time encoding strategies, covering training-free and fine-tuned settings on three complementary video understanding tasks. Our results reveal task-dependent compression behavior and characterize the effects of compression placement, time encoding, and backbone choice. These findings provide empirical guidance for designing and selecting video token compression strategies across tasks and training settings.

제출 이력

확인된 버전만 표시합니다. 전체 제출 이력은 arXiv 원문에서 확인하세요.

  1. arXiv:2611.25574v1제출 시각 미확인

함께 살펴볼 논문

공통 태그를 근거로 연결했습니다. 실제 인용 관계나 논문의 우수성을 뜻하지 않습니다.

공통 태그로 연결된 공개 논문이 아직 없습니다.

토론

이 논문과 연결된 포럼 글이 아직 없습니다.