Ar2Can：以建筑师与画家借画布实现多人图像生成
Ar2Can: An Architect and an Artist Leveraging a Canvas for Multi-Human Generation
把多人生成拆成「布局规划 + 身份渲染」两阶段，并用匈牙利匹配的空间对齐人脸匹配奖励配合 GRPO 优化，可直接迁移到任何需要多主体身份保持的扩散生成/编辑流程中。
原文依据 a spatially-grounded face matching reward that combines Hungarian spatial alignment with identity similarity
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论尽管个性化图像生成近期取得进展，现有模型始终无法可靠地生成多人场景，常常出现面部身份融合或丢失。我们提出 Ar2Can，一个新颖的两阶段框架，将多人生成中的空间规划与身份渲染解耦。建筑师（Architect）预测结构化布局，指定每个人应出现的位置；画家（Artist）随后合成照片级真实图像，并由一个空间接地的人脸匹配奖励引导，该奖励将匈牙利空间对齐与身份相似度相结合。这一做法确保人脸被渲染在正确位置并忠实保留参考身份。我们开发了两种建筑师变体，并与基于扩散的画家模型无缝集成。该模型通过组相对策略优化（GRPO）进行优化，使用针对数量准确性、图像质量和身份匹配的组合奖励。在 MultiHuman-Testbench 上评估，Ar2Can 在数量准确性和身份保持两方面均取得显著提升，同时保持高感知质量。值得注意的是，我们的方法主要使用合成数据即取得这些结果，无需真实多人图像。项目主页：https://qualcomm-ai-research.github.io/ar2can/
英文原文摘要 1270 字符
Despite recent advances in personalized image generation, existing models consistently fail to produce reliable multi-human scenes, often merging or losing facial identity. We present Ar2Can, a novel two-stage framework that disentangles spatial planning from identity rendering for multi-human generation. The Architect predicts structured layouts, specifying where each person should appear. The Artist then synthesizes photorealistic images, guided by a spatially-grounded face matching reward that combines Hungarian spatial alignment with identity similarity. This approach ensures faces are rendered at correct locations and faithfully preserve reference identities. We develop two Architect variants, seamlessly integrated with our diffusion-based Artist model. This is optimized via Group Relative Policy Optimization (GRPO) using compositional rewards for count accuracy, image quality, and identity matching. Evaluated on the MultiHuman-Testbench, Ar2Can achieves substantial improvements in both count accuracy and identity preservation, while maintaining high perceptual quality. Notably, our method achieves these results using primarily synthetic data, without requiring real multi-human images. Project page: https://qualcomm-ai-research.github.io/ar2can/