通过神经场适配进行权重空间表征学习
Weight Space Representation Learning via Neural Field Adaptation
把预训练神经场的乘性低秩适配权重本身当作图像/3D 场景的表征，可直接接入潜变量扩散模型做生成，为视觉研究提供了一条绕开像素或特征编码器的「权重即表征」路线，可用于少样本重建与权重空间编辑。
原文依据 multiplicative LoRA weights achieve high representation quality while exhibiting distinctiveness and semantic structure
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论本文探讨权重作为有效表征的潜力，聚焦于神经场。我们的核心洞见是：通过预训练基模型与低秩适配（LoRA）约束优化空间，可以在权重空间中诱导出结构。在面向 2D 与 3D 数据的重建、生成与分析任务中，我们发现乘性 LoRA 权重在取得高表征质量的同时，还表现出区分性与语义结构。当与潜变量扩散模型结合使用时，乘性 LoRA 权重能够实现比现有权重空间方法更高质量的生成。
英文原文摘要 625 字符
In this work, we investigate the potential of weights to serve as effective representations, focusing on neural fields. Our key insight is that constraining the optimization space through a pre-trained base model and low-rank adaptation (LoRA) can induce structure in weight space. Across reconstruction, generation, and analysis tasks on 2D and 3D data, we find that multiplicative LoRA weights achieve high representation quality while exhibiting distinctiveness and semantic structure. When used with latent diffusion models, multiplicative LoRA weights enable higher-quality generation than existing weight-space methods.