ORION：面向通用 VLM 适配的正交文本编码
ORION: ORthonormal Text Encoding for Universal VLM AdaptatION
把文本编码器的类别原型当作可优化对象，用低秩适配加正交性损失在仅有类名的条件下重塑 CLIP 文本嵌入，可直接替换下游零样本/少样本/测试时自适应流程中的手工 prompt 原型，是一种即插即用的分类头改进手段。
原文依据 We introduce ORION, a text encoder fine-tuning framework that improves pretrained VLMs using only class names.
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论视觉语言模型（VLM）在多种任务上展现出显著的泛化能力，但其性能仍受限于用于表示类别的文本原型的质量与几何结构。由冻结的文本编码器和手工设计的提示词得到的标准零样本分类器，可能产生相关性强或区分度弱的嵌入，从而限制任务特定的判别能力。我们提出 ORION，一个仅使用类别名称即可改进预训练 VLM 的文本编码器微调框架。该方法通过低秩适配优化一个由两项组成的新损失：一项促进给定任务各类别文本表示之间的两两正交性，另一项惩罚对初始类别原型的偏离。此外，我们给出了正交性惩罚的概率解释，借助惠更斯定理将其与通用的最大似然估计（MLE）原理联系起来。我们在 11 个基准和三个大型 VLM 骨干网络上开展了大量实验，表明精炼后的文本嵌入可作为标准 CLIP 原型的有力替代。作为即插即用模块叠加在各种最优方法之上，并在不同预测设定（零样本、少样本和测试时自适应）下，ORION 都能持续且显著地提升性能。
英文原文摘要 1367 字符
Vision-language models (VLMs) have demonstrated remarkable generalization across diverse tasks, yet their performance remains constrained by the quality and geometry of the textual prototypes used to represent classes. Standard zero-shot classifiers, derived from frozen text encoders and handcrafted prompts, may yield correlated or weakly separated embeddings that limit task-specific discriminability. We introduce ORION, a text encoder fine-tuning framework that improves pretrained VLMs using only class names. Our method optimizes, via low-rank adaptation, a novel loss integrating two terms, one promoting pairwise orthogonality between the textual representations of the classes of a given task and the other penalizing deviations from the initial class prototypes. Furthermore, we provide a probabilistic interpretation of our orthogonality penalty, connecting it to the general maximum likelihood estimation (MLE) principle via Huygens' theorem. We report extensive experiments on 11 benchmarks and three large VLM backbones, showing that the refined textual embeddings yield powerful replacements for the standard CLIP prototypes. Added as plug-and-play module on top of various state-of-the-art methods, and across different prediction settings (zero-shot, few-shot and test-time adaptation), ORION improves the performance consistently and significantly.