提出Circle-RoPE，解决视觉语言模型中位置编码的跨模态干扰问题。
Circle-RoPE: Cone-like Decoupled Rotary Positional Embedding for Large Vision-Language Models
- 将图像位置映射到环形结构，实现图文位置解耦
- 在多个基准上提升空间定位与视觉推理能力
- 适合需要精准空间理解的多模态任务
旋转位置编码（RoPE）广泛应用于大语言模型，但用于视觉语言模型（VLM）时会耦合文本与图像的位置索引，引入虚假的跨模态相对位置偏差。本文提出每标记距离（PTD）来量化跨模态位置解耦程度，并证明PTD = 0是消除由RoPE引起的几何注意力偏差的充分条件。基于此，我们提出Circle-RoPE，将2D图像标记坐标映射到与文本位置轴正交的环形区域，形成锥状几何结构，使每个文本标记到所有图像标记的距离相等，同时保留图像内部的空间结构。进一步提出交替几何编码（AGE），通过在不同层间交替使用Circle-RoPE的解耦几何与标准RoPE的网格先验，融合互补的几何信息。实验表明，在多种VLM主干网络和多模态基准上均取得一致的性能提升，尤其在空间定位与视觉推理方面表现优异。代码已开源：https://github.com/lose4578/CircleRoPE。
原文摘要 · Abstract (English)
Rotary Position Embedding (RoPE) is widely adopted in large language models, but when applied to vision-language models (VLMs) it couples text and image position indices and can introduce spurious cross-modal relative-position bias. We propose Per-Token Distance (PTD) to quantify cross-modal positional disentanglement, and prove that PTD = 0 is a sufficient condition to eliminate the geometric attention bias induced by RoPE. Guided by this criterion, we introduce Circle-RoPE, which remaps 2D image-token coordinates onto an annulus orthogonal to the text position axis, yielding a cone-like geometry where each text token is equidistant to all image tokens while preserving intra-image spatial structure. We further propose Alternating Geometry Encoding (AGE) to combine complementary geometric priors by alternating the decoupled geometry of Circle-RoPE and the grid-based prior of standard RoPE across layers. This design enables cross-modal positional disentanglement while preserving fine-grained intra-image spatial structure. Experiments on diverse VLM backbones and multimodal benchmarks show consistent gains in spatial grounding and visual reasoning. The code is available at https://github.com/lose4578/CircleRoPE.
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。