arXiv:2608.08630cs.CV2026-08

VLZip通过统一压缩图文序列,实现超长上下文下的高保真多模态推理。

VLZip: Unified Visual and Textual Compression for Interleaved Long-Context Modeling

论文配图:VLZip: Unified Visual and Textual Compression for Interleaved Long-Context Modeling
图 1 · 摘自论文原文
  • 将图文内容分层压缩为轻量级软前缀注入解码层
  • 支持120K训练tokens(基线6倍)和280K以上推理,内存显著降低
  • 适用于视频叙事等需全局理解的长序列多模态任务

视觉语言模型在处理超长交错图文序列时面临自注意力计算复杂度平方增长的问题。现有方法或过度裁剪令牌导致信息丢失,或采用低精度高效架构,且普遍忽视文本组件的重要性。我们提出VLZip框架,通过统一压缩视觉与文本内容,在纯Transformer结构中实现高保真推理。核心思想是将图文段分层提炼为层特定的紧凑‘软前缀’,并注入每个解码层隐藏状态,大幅缩短注意力序列同时保留细粒度全局上下文。为弥补领域评估不足,我们还构建了来自视频叙事的LongVLBench新基准,要求整体叙事级推理。大量实验表明,VLZip在长上下文多模态推理上表现领先,可支持最高120K tokens训练(较基线提升6倍),推理超过280K tokens且内存显著减少,并具备扩展至200万tokens的内存可扩展性。在现有方法崩溃的极端上下文长度下表现出色,确立了长上下文多模态AI的新高效标准。代码已开源:https://github.com/ShareLab-SII/VLZip。

原文摘要 · Abstract (English)

Vision Language Models (VLMs) face significant challenges with ultra-long, interleaved image-text sequences due to the quadratic complexity of self-attention. Current solutions either resort to aggressive token pruning, risking irreversible information loss, or adopt efficient but less precise architectures, while largely ignoring the equally vital textual component. We introduce VLZip, a framework that unifies visual and textual compression for high-fidelity reasoning within a pure Transformer. At its core, VLZip hierarchically distills visual and textual segments into compact, layer-specific "soft prefixes" and injects them into each decoder layer's hidden states, drastically shortening the attention sequence while preserving fine-grained global context. To address deficient evaluations in the field, we also introduce LongVLBench, a new benchmark derived from video narratives that demands holistic, narrative-level reasoning. Extensive experiments show VLZip achieves leading performance on long-context multimodal reasoning, enabling training up to 120K tokens, a 6x increase over the baseline, and inference beyond 280K tokens with significantly reduced memory, while demonstrating the memory scalability to handle up to 2M tokens. By excelling at extreme context lengths where existing methods collapse, VLZip establishes an efficient and powerful new standard for long-context multimodal AI. Code is available at https://github.com/ShareLab-SII/VLZip.

多模态长序列压缩Transformer

Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品;中文卡片由大模型生成,请以原文为准。