提出新方法解码ViT内部机制，看清模块间信息流动。
Decoding Vision Transformers: the Diffusion Steering Lens
- 通过引导子模块输出，追踪视觉变换器中信息传递路径。
- 干预实验验证该方法能准确揭示各模块对最终结果的间接贡献。
- 无需训练，适合研究ViT可解释性与模型调试的学者使用。
Logit Lens 是一种广泛用于基于Transformer的语言模型机制可解释性的方法，可通过将内部表示投影到输出词汇空间来分析其在各层间的演变。尽管将Logit Lens直接应用于视觉变换器（ViTs）在技术上是可行的，但其在捕捉视觉表征丰富性方面存在局限。基于Toker等人（2024）提出的扩散镜头（Diffusion Lens），该方法可有效可视化文本到图像扩散模型中编码器的中间表示，但无法捕捉单个子模块的直接贡献。为克服这一限制，我们提出 extbf{扩散引导镜头}（Diffusion Steering Lens, DSL），一种无需训练的新方法，通过引导子模块输出并追踪其后续的间接影响。我们通过干预研究验证了该方法，结果表明DSL能直观且可靠地解释ViTs中的内部处理过程。
原文摘要 · Abstract (English)
Logit Lens is a widely adopted method for mechanistic interpretability of transformer-based language models, enabling the analysis of how internal representations evolve across layers by projecting them into the output vocabulary space. Although applying Logit Lens to Vision Transformers (ViTs) is technically straightforward, its direct use faces limitations in capturing the richness of visual representations. Building on the work of Toker et al. (2024)~\cite{Toker2024-ve}, who introduced Diffusion Lens to visualize intermediate representations in the text encoders of text-to-image diffusion models, we demonstrate that while Diffusion Lens can effectively visualize residual stream representations in image encoders, it fails to capture the direct contributions of individual submodules. To overcome this limitation, we propose \textbf{Diffusion Steering Lens} (DSL), a novel, training-free approach that steers submodule outputs and patches subsequent indirect contributions. We validate our method through interventional studies, showing that DSL provides an intuitive and reliable interpretation of the internal processing in ViTs.
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。