CARLoS：通过 LoRA 的简洁评估表示实现大规模检索
CARLoS: Retrieval via Concise Assessment Representation of LoRAs at Scale
用 CLIP 嵌入与基模型生成结果的差值构造 Direction/Strength/Consistency 三段式表示，为 LoRA 这类生成组件的无元数据刻画与检索提供了可复用范式，可直接迁移到视觉生成模型的适配器管理与组合检索。
原文依据 Using CLIP embeddings and their difference to a base-model generation, we concisely define a three-part representation
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论生成组件（如 LoRA）的迅速激增造就了一个庞大但缺乏结构化的生态。现有的发现方法依赖不可靠的用户描述或有偏的热度指标，损害了可用性。我们提出 CARLoS，一个无需额外元数据即可刻画 LoRA 的大规模框架。在分析超过 650 个 LoRA 的基础上，我们在多种提示与随机种子上使用它们进行图像生成，以此作为评估其行为的可信方式。利用 CLIP 嵌入及其与基模型生成结果的差异，我们简洁地定义了一种三部分表示：方向（Direction），刻画语义偏移；强度（Strength），量化影响的显著程度；一致性（Consistency），量化该影响的稳定程度。基于这些表示，我们开发了一个高效的检索框架，可将文本查询语义匹配到相关 LoRA，同时过滤掉过强或不稳定的 LoRA，在自动化与人工评估中均优于文本基线。虽然检索是我们的主要关注点，但同一表示还支持将强度与一致性同实质性（substantiality）与意志性（volition）的法律概念相联系的分析，这两者是版权中的关键考量，从而使 CARLoS 成为一个具有更广泛 LoRA 分析相关性的实用系统。
英文原文摘要 1274 字符
The rapid proliferation of generative components, such as LoRAs, has created a vast but unstructured ecosystem. Existing discovery methods depend on unreliable user descriptions or biased popularity metrics, hindering usability. We present CARLoS, a large-scale framework for characterizing LoRAs without requiring additional metadata. Analyzing over 650 LoRAs, we employ them in image generation over a variety of prompts and seeds, as a credible way to assess their behavior. Using CLIP embeddings and their difference to a base-model generation, we concisely define a three-part representation: Direction, defining semantic shift; Strength, quantifying the significance of the effect; and Consistency, quantifying how stable the effect is. Using these representations, we develop an efficient retrieval framework that semantically matches textual queries to relevant LoRAs while filtering overly strong or unstable ones, outperforming textual baselines in automated and human evaluations. While retrieval is our primary focus, the same representation also supports analyses linking Strength and Consistency to legal notions of substantiality and volition, key considerations in copyright, positioning CARLoS as a practical system with broader relevance for LoRA analysis.