# The Independence Prior of SAEs Fragments Visual Concepts

> 全文为英语

**Published:** 2026年10月6日
**Author:** Tommaso Mencattini, Giorgos Nikolaou, Donato Crisostomi, Thomas Fel, Francesco Montagna, Emanuele Rodol\`a, Francesco Locatello
**Source:** arXiv — cs.AI
**Language:** en-US
**HTML:** https://aiqazaqstan.com/zh/articles/the-independence-prior-of-saes-fragments-visual-concepts

Sparse Autoencoders (SAEs) decompose model activations into sparse combinations of interpretable dictionary atoms. Although SAEs are grounded in the Linear Representation Hypothesis (LRH), their objective smuggles in an additional prior: concepts across patches are treated as…

> 内容来自原始来源： **arXiv — cs.AI**
> [在 arXiv — cs.AI 阅读全文](https://arxiv.org/abs/2610.04112)

**标签：** #arXiv

---

*AI Qazaqstan — non-commercial informational portal about AI in Kazakhstan. Machine translations and summaries are AI-generated.*