发表机构
University of Nottingham; Lancaster University; Peking University（诺丁汉大学; 兰卡斯特大学; 北京大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
本文提出无训练的CANVAS框架，通过视觉自适应采样提升长上下文文本到SVG生成的全局一致性，无需额外训练即可在多基准上取得良好效果。
AI 中文摘要
自回归大模型近期推动文本到SVG生成从简单图标扩展到复杂长上下文图形，但标准自回归解码常无法保持几何、布局、遮挡及构图间的全局一致性。本文提出CANVAS（Consistency-Aware Navigation via Visual Adaptive Sampling，即一致性感知视觉自适应采样导航），这是一种无训练、感知渲染的推理框架，结合锐化轨迹似然与渲染未来的视觉反馈，推导逐笔画导航规则。该框架在有限生成与渲染预算下有效估计每一条候选笔画的未来价值，并根据候选不确定性、决策影响力及展开成本自适应分配采样。在多个自回归SVG主干及互补基准上开展的实验表明，该框架无需额外训练即可提升全局一致性，包括合理几何关系、空间布局、遮挡顺序及整体构图，验证了其有效性与泛化能力。
英文摘要
Autoregressive large models have recently advanced Text-to-SVG generation from simple icons to complex, long-context graphics, yet standard autoregressive decoding often fails to maintain global consistency across geometry, layout, occlusion, and composition. We introduce CANVAS (Consistency-Aware Navigation via Visual Adaptive Sampling), a training-free, render-aware inference framework that combines power-sharpened trajectory likelihood with visual feedback from rendered futures and derives a stroke-wise navigation rule. It effectively estimates each candidate stroke's future value under a limited generation and rendering budget and adaptively allocates samples according to candidate uncertainty, decision influence, and rollout cost. Experiments across multiple autoregressive SVG backbones and complementary benchmarks demonstrate improvements in global consistency, which includes sound geometric relationships, spatial layouts, occlusion ordering, and overall composition, without additional training, demonstrating the effectiveness and generalization ability of our framework.
Comments30 pages (including supplementary material), 9 figures, 10 tables. Code: https://github.com/Louis-YW/CANVAS
