PARCEL: 基于池锚定的条件弹性查询重采样以实现高效视觉-语言理解
PARCEL: Pool-Anchored Resampling with Conditioned Elastic Queries for Efficient Vision-Language Understanding
AI总结：
提出PARCEL视觉分词架构，通过池锚定和条件弹性查询重采样解决视觉令牌压缩中的空间与查询表示冲突，在27个基准上提升性能-效率帕累托前沿。
摘要：
大型视觉-语言模型（LVLMs）将视觉输入映射为密集的令牌序列，导致推理时的二次计算瓶颈。弹性视觉令牌压缩通过训练单一模型以在多个视觉令牌预算下运行来解决这一问题。然而，现有方法在激进压缩下表现不佳。空间压缩（如嵌套池化）表现为不完美的低通滤波器，并引起频谱混叠，掩盖了细粒度细节。查询压缩（如嵌套查询重采样）用非局部摘要替代显式的网格对齐令牌，显著降低了空间定位能力。为解决这一表示冲突，我们引入了PARCEL（基于池锚定的条件弹性查询重采样以实现高效视觉-语言理解），一种视觉分词架构，动态分配特征提取的工作。PARCEL将空间池令牌建立为低频布局锚点，并通过池条件查询重采样使弹性查询令牌依赖于这些锚点。这鼓励查询令牌专注于互补的视觉特征，而非冗余的空间映射。在27个基准上的广泛评估表明，PARCEL改进了性能-效率帕累托前沿，在各种视觉令牌预算下持续优于现有的嵌套基线，同时保留了“一次训练，随处部署”的范式。
英文摘要：
Large Vision-Language Models (LVLMs) map visual inputs into dense token sequences, imposing a quadratic computational bottleneck for inference. Elastic visual-token compression addresses this by training a single model that can run at multiple visual-token budgets. However, existing approaches struggle under aggressive compression. Spatial-only compression, as in nested pooling, behaves as an imperfect low-pass filter and induces spectral aliasing that obscures fine-grained detail. Query-only compression, as in nested query resampling, replaces explicit grid-aligned tokens with non-local summaries and substantially degrades spatial grounding. To resolve this representational conflict, we introduce PARCEL (Pool-Anchored Resampling with Conditioned Elastic Queries for Efficient Vision-Language Understanding), a visual tokenization architecture that dynamically partitions the labor of feature extraction. PARCEL establishes spatial pool tokens as low-frequency layout anchors and conditions elastic query tokens on these anchors through Pool-Conditioned Query Resampling. This encourages query tokens to focus on complementary visual features rather than redundant spatial mapping. Extensive evaluations across 27 benchmarks show that PARCEL improves the performance-efficiency Pareto frontier, consistently outperforming existing matryoshka baselines across visual-token budgets while preserving the "train once, deploy anywhere" paradigm.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
