← 研踪 PaperTrailarXiv:2611.05522 · v1
From Activation to Intervention: Interpreting Sparse Visual Features through Reconstruction
首次提交 2026-11-04 03:17
版本更新 2026-11-04 03:17
雷达观测 2026-11-05 10:08
摘要
Sparse autoencoders (SAEs) decompose visual representations into individual features, offering a promising basis for explaining model behavior. Interpreting these features typically relies on images or pixel regions that most strongly activate them. However, this evidence becomes ambiguous when multiple features with different semantics frequently activate in the same locations, such as texture and color in the same patterns. We propose Reconstructive SAE Intervention (RecSAEI), a framework that shifts feature explanation from passive observation to active visual reconstruction. By intervening on representations along SAE decoder directions and reconstructing the visual output, we provide direct evidence of a feature's semantic meaning. We demonstrate that our method successfully disambiguates feature meanings across multiple datasets and vision encoders. Furthermore, our framework grounds [CLS] SAE interventions in a sparse set of patch SAE decoder directions. Beyond stable target response, this attribution also exposes the compositional relationship between elementary visual patterns and high-level semantics. Our results advocate for intervention reconstruction as an essential complementary signal for explaining neural networks.
作者备注：Demo and code available at https://imageomics.github.io/RecSAEI
观测到的版本
相邻记录可能跳过中间版本；首次发现 v2 及以上也会标为版本更新。
v1 · 2026-11-04 03:17 · 最新观测版本
From Activation to Intervention: Interpreting Sparse Visual Features through Reconstruction
Sparse autoencoders (SAEs) decompose visual representations into individual features, offering a promising basis for explaining model behavior. Interpreting these features typically relies on images or pixel regions that most strongly activate them. However, this evidence becomes ambiguous when multiple features with different semantics frequently activate in the same locations, such as texture and color in the same patterns. We propose Reconstructive SAE Intervention (RecSAEI), a framework that shifts feature explanation from passive observation to active visual reconstruction. By intervening on representations along SAE decoder directions and reconstructing the visual output, we provide direct evidence of a feature's semantic meaning. We demonstrate that our method successfully disambiguates feature meanings across multiple datasets and vision encoders. Furthermore, our framework grounds [CLS] SAE interventions in a sparse set of patch SAE decoder directions. Beyond stable target response, this attribution also exposes the compositional relationship between elementary visual patterns and high-level semantics. Our results advocate for intervention reconstruction as an essential complementary signal for explaining neural networks.
查看此版本解读 →查看此版本原文 ↗