让不同模型共享同一组概念空间，实现跨模型跨模态解释性对齐。
SPARC: Concept-Aligned Sparse Autoencoders for Cross-Model and Cross-Modal Interpretability
- 通过全局TopK稀疏机制和跨重建损失，统一多模型概念表示。
- 在Open Images上实现0.80的雅可比相似度，对齐效果提升超三倍。
- 适合需要对比不同模型如何表征相同概念的研究者使用。
理解不同AI模型如何编码相同高层概念（如物体或属性）仍具挑战性，因各模型通常生成独立表示。现有可解释性方法如稀疏自编码器（SAEs）为每个模型单独生成潜在概念，导致概念空间不兼容，限制跨模型可解释性。为此，我们提出SPARC（概念对齐的稀疏自编码器），一种新框架，在多种架构与模态间学习单一统一的潜在空间（如视觉模型DINO、多模态模型CLIP）。SPARC的对齐通过两项关键创新实现：(1) 全局TopK稀疏机制，确保所有输入流在特定概念下激活相同的潜在维度；(2) 跨重建损失，显式促进模型间语义一致性。在Open Images数据集上，SPARC显著提升概念对齐，达到0.80的雅可比相似度，较此前方法提升超过三倍。SPARC构建了一个共享稀疏潜在空间，其中各维度常对应跨模型与模态的相似高层概念，实现无需人工对齐即可直接比较不同架构对同一概念的表征。由此产生的对齐表示还可支持实际应用，如在纯视觉模型中实现文本引导的空间定位，以及跨模型/跨模态检索。代码与模型已开源于https://github.com/AtlasAnalyticsLab/SPARC。
原文摘要 · Abstract (English)
Understanding how different AI models encode the same high-level concepts, such as objects or attributes, remains challenging because each model typically produces its own isolated representation. Existing interpretability methods like Sparse Autoencoders (SAEs) produce latent concepts individually for each model, resulting in incompatible concept spaces and limiting cross-model interpretability. To address this, we introduce SPARC (Sparse Autoencoders for Aligned Representation of Concepts), a new framework that learns a single, unified latent space shared across diverse architectures and modalities (e.g., vision models like DINO, and multimodal models like CLIP). SPARC's alignment is enforced through two key innovations: (1) a Global TopK sparsity mechanism, ensuring all input streams activate identical latent dimensions for a given concept; and (2) a Cross-Reconstruction Loss, which explicitly encourages semantic consistency between models. On Open Images, SPARC dramatically improves concept alignment, achieving a Jaccard similarity of 0.80, more than tripling the alignment compared to previous methods. SPARC creates a shared sparse latent space where individual dimensions often correspond to similar high-level concepts across models and modalities, enabling direct comparison of how different architectures represent identical concepts without requiring manual alignment or model-specific analysis. As a consequence of this aligned representation, SPARC also enables practical applications such as text-guided spatial localization in vision-only models and cross-model/cross-modal retrieval. Code and models are available at https://github.com/AtlasAnalyticsLab/SPARC
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。