arXiv：cs.AI· Karim Zaghw, Wieland Brendel, Patrik Reizinger·· 22 小时前
学习到的块何时保留因果含义？块可识别性与构造性因果抽象
When Do Learned Blocks Preserve Causal Meaning? Block Identifiability and Constructive Causal Abstractions
AI 导读
研究将块可识别性与构造性因果抽象相连接，证明块对角可逆变换加兼容坐标置换是保留基于这些块构建的每一个构造性因果抽象的最大变换类，该保证的前提是块与概念的对应关系随表示更新。在加性自编码器上的小型合成实验显示，高层干预可在独立学习到的表示之间迁移，并近似保留其预期效果。
正文
Abstract:Standard identifiability theory assumes one-dimensional latent factors, but learned representations can encode latent information across multidimensional blocks whose internal coordinates are not uniquely determined. When such blocks are used to represent and intervene on higher-level concepts, it is unclear whether their causal meaning survives the transformations left unresolved by block-identifiability. We connect block-identifiability to constructive causal abstraction. We show that block-diagonal invertible transformations followed by compatible coordinate permutations form the maximal class that preserves every constructive causal abstraction built on those blocks. This guarantee assumes that the correspondence between blocks and concepts is updated with the representation. A small synthetic experiment with additive autoencoders shows that high-level interventions can be transferred between independently learned representations while approximately preserving their intended effects.
| Subjects: | Machine Learning (cs.LG); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2610.18155 [cs.LG] |
| (or arXiv:2610.18155v1 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2610.18155 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Karim Zaghw [view email]
[v1]
Tue, 13 Oct 2026 11:00:57 UTC (90 KB)
来源：arXiv：cs.AI · arxiv.org