MultiCrafter：通过解耦注意力与身份感知偏好对齐实现高保真多主体生成
MultiCrafter: High-Fidelity Multi-Subject Generation via Disentangled Attention and Identity-Aware Preference Alignment
提出两阶段解耦训练框架：预训练阶段用显式位置监督解决注意力渗漏，后训练阶段用基于匈牙利匹配的在线强化学习做身份保持的偏好优化，可直接用于多主体个性化生成任务。
原文依据 we introduce an explicit positional supervision mechanism that effectively resolves attention bleeding and drastically enhances subject fidelity
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论多主体图像生成旨在在单张图像中合成用户提供的多个主体，同时保持主体保真度、确保提示词一致性并符合人类审美偏好。现有的基于上下文学习的方法受限于其高度耦合的训练范式。这些方法试图在单一训练阶段同时实现高主体保真度和多维人类偏好对齐，依赖单一且间接的重建损失，难以同时满足这两个目标。为此，我们提出 MultiCrafter，一个将该任务解耦为两个独立训练阶段的框架。首先，在预训练阶段，我们引入显式位置监督机制，有效解决注意力渗漏问题并大幅提升主体保真度。其次，在后训练阶段，我们提出身份保持偏好优化，一种新颖的在线强化学习框架。我们设计了一种基于匈牙利匹配算法的评分机制来准确评估多主体保真度，使模型能够在优化美学和提示词对齐的同时，确保第一阶段获得的主体保真度。实验验证表明，我们的解耦框架显著提升了主体保真度，并更好地与人类偏好对齐。
英文原文摘要 1372 字符
Multi-subject image generation aims to synthesize user-provided subjects in a single image while preserving subject fidelity, ensuring prompt consistency, and aligning with human aesthetic preferences. Existing In-Context-Learning based methods are limited by their highly coupled training paradigm. These methods attempt to achieve both high subject fidelity and multi-dimensional human preference alignment within a single training stage, relying on a single, indirect reconstruction loss, which is difficult to simultaneously satisfy both these goals. To address this, we propose MultiCrafter, a framework that decouples this task into two distinct training stages. First, in a pre-training stage, we introduce an explicit positional supervision mechanism that effectively resolves attention bleeding and drastically enhances subject fidelity. Second, in a post-training stage, we propose Identity-Preserving Preference Optimization, a novel online reinforcement learning framework. We feature a scoring mechanism to accurately assess multi-subject fidelity based on the Hungarian matching algorithm, which allows the model to optimize for aesthetics and prompt alignment while ensuring subject fidelity achieved in the first stage. Experiments validate that our decoupling framework significantly improves subject fidelity while aligning with human preferences better.