LoFA：学习预测个性化先验以实现视觉生成模型的快速适配
LoFA: Learning to Predict Personalized Prior for Fast Adaptation of Visual Generative Models
把 LoRA 微调从「逐任务优化数小时」变成「超网络秒级预测」，视觉生成模型做个性化时可直接复用其两阶段超网络（先预测相对分布模式、再回归 LoRA 权重）作为免优化适配器。
原文依据 we design a two-stage hypernetwork: first predicting relative distribution patterns that capture key adaptation regions, then using these to guide final LoRA weight prediction
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论将视觉生成模型个性化以满足特定用户需求已受到越来越多的关注，然而当前诸如低秩适配（LoRA）之类的方法由于需要任务特定数据和冗长的优化过程而仍不实用。尽管少数基于超网络的方法尝试直接预测适配权重，但它们难以将细粒度的用户提示映射到复杂的 LoRA 分布，限制了其实际适用性。为弥合这一差距，我们提出 LoFA，一个能够高效预测个性化先验以实现模型快速适配的通用框架。我们首先识别出 LoRA 的一个关键性质：在 LoRA 与基础模型参数的相对变化中会涌现出结构化的分布模式。基于此，我们设计了一个两阶段超网络：先预测刻画关键适配区域的相对分布模式，再利用这些模式指导最终的 LoRA 权重预测。大量实验表明，我们的方法能够在数秒内跨多个任务和用户提示一致地预测出高质量的个性化先验，甚至优于需要数小时处理的传统 LoRA。
英文原文摘要 1151 字符
Personalizing visual generative models to meet specific user needs has gained increasing attention, yet current methods like Low-Rank Adaptation (LoRA) remain impractical due to their demand for task-specific data and lengthy optimization. While a few hypernetwork-based approaches attempt to predict adaptation weights directly, they struggle to map fine-grained user prompts to complex LoRA distributions, limiting their practical applicability. To bridge this gap, we propose LoFA, a general framework that efficiently predicts personalized priors for fast model adaptation. We first identify a key property of LoRA: structured distribution patterns emerge in the relative changes between LoRA and base model parameters. Building on this, we design a two-stage hypernetwork: first predicting relative distribution patterns that capture key adaptation regions, then using these to guide final LoRA weight prediction. Extensive experiments demonstrate that our method consistently predicts high-quality personalized priors within seconds, across multiple tasks and user prompts, even outperforming conventional LoRA that requires hours of processing.