← 研踪 PaperTrailarXiv:2608.28008 · v2
Visual Token Coding for Video Multimodal Large Language Models
首次提交 2026-08-28 15:21
版本更新 2026-11-18 21:20
雷达观测 2026-11-19 15:04
摘要
In this paper, we propose a new token compression paradigm for video Multimodal Large Language Models (MLLMs), termed Visual Token Coding (). Inspired by classical video coding principles, e.g., HEVC, VTC performs structured compression by predicting the I/P frames of a video and measuring their frame-wise residuals to estimate token redundancy. Based on this base paradigm, we further extend VTC with a set of dynamic budgeting designs towards more cost-effective token coding, termed . To validate VTC and , we apply them to three MLLMs and conduct experiments on a set of video understanding benchmarks. The experimental results show that as a base paradigm, VTC can achieve an average performance retention of 98.4% on Qwen3-VL under a 50% token compression ratio, and the enhanced can further improve the retention to 100.3%, outperforming a set of compared methods. Moreover, as a plug-and-play design, VTC requires no additional tuning of MLLMs for token coding. Our code is given in the supplementary materials.
作者备注:11 pages, 5 figures
观测到的版本
相邻记录可能跳过中间版本;首次发现 v2 及以上也会标为版本更新。
v2 · 2026-11-18 21:20 · 最新观测版本
Visual Token Coding for Video Multimodal Large Language Models
In this paper, we propose a new token compression paradigm for video Multimodal Large Language Models (MLLMs), termed Visual Token Coding (). Inspired by classical video coding principles, e.g., HEVC, VTC performs structured compression by predicting the I/P frames of a video and measuring their frame-wise residuals to estimate token redundancy. Based on this base paradigm, we further extend VTC with a set of dynamic budgeting designs towards more cost-effective token coding, termed . To validate VTC and , we apply them to three MLLMs and conduct experiments on a set of video understanding benchmarks. The experimental results show that as a base paradigm, VTC can achieve an average performance retention of 98.4% on Qwen3-VL under a 50% token compression ratio, and the enhanced can further improve the retention to 100.3%, outperforming a set of compared methods. Moreover, as a plug-and-play design, VTC requires no additional tuning of MLLMs for token coding. Our code is given in the supplementary materials.
查看此版本解读 →查看此版本原文 ↗