DCoAR：面向个性化文本到图像生成的统一自回归模型深度概念注入
DCoAR: Deep Concept Injection into Unified Autoregressive Models for Personalized Text-to-Image Generation
对做个性化/定制化生成的人，这篇给出了在完全冻结的统一自回归模型上做深层概念注入的路线：逐层多模态上下文学习 + 双重先验保持与上下文感知自正则化，可在不微调主干的前提下实现主体定制与用户风格迁移，是替代适配式微调的低参数方案。
原文依据 a novel deep concept injection framework that maintains a completely frozen pre-trained model
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论统一自回归（AR）模型在多模态理解与生成方面表现优异，但其在定制化图像生成领域的全部潜力尚未被充分发掘。现有的统一 AR 模型定制方法面临一个根本性困境：基于适配的方法受困于过拟合与可扩展性瓶颈，而概念注入范式则受限于浅层注入策略，导致视觉保真度不佳且重上下文化能力受损。为解决这一问题，我们提出 DCoAR，一种全新的深度概念注入框架，其预训练模型保持完全冻结。DCoAR 通过逐层多模态上下文学习（LMCL）策略深度整合新概念，并由多方面的正则化方案加以稳定：双重先验保持（DPP）损失用于缓解语义漂移，上下文感知自正则化（CASR）损失用于增强重上下文化能力。该框架还支持在用户提供的风格下进行免训练的主体定制。实验表明，DCoAR 显著优于以往的注入式方法，并在可训练参数大幅减少的情况下取得与基于适配的方法相竞争的性能。
英文原文摘要 1256 字符
The unified autoregressive (AR) model excels at multimodal understanding and generation. However, its full potential in the domain of customized image generation has yet to be fully realized.Existing customization approaches for unified AR models face a fundamental dilemma: adaptation-based methods suffer from overfitting and scalability bottlenecks, while concept-injection paradigms are constrained by a shallow injection strategy that leads to poor visual fidelity and impaired re-contextualization.To address this, we propose DCoAR, a novel deep concept injection framework that maintains a completely frozen pre-trained model. DCoAR deeply integrates new concepts through a Layer-wise Multimodal Context Learning (LMCL) strategy, which is stabilized by a multi-faceted regularization scheme: a Dual Prior Preservation (DPP) loss to mitigate semantic drift and a Context-Aware Self-Regularization (CASR) loss to enhance re-contextualization. The framework also enables training-free subject customization in user-provided styles.Experiments demonstrate that DCoAR significantly outperforms previous injection-based methods and achieves performance competitive with adaptation-based approaches while requiring substantially fewer trainable parameters.