# Google's Sparse Autoencoder

**URL:** <https://discourse.numenta.org/t/googles-sparse-autoencoder/11439>\
**Category:** Related Papers\
**Created:** [August 3, 2024, 12:42pm UTC](https://discourse.numenta.org/t/googles-sparse-autoencoder/11439 "2024-08-03T12:42:41Z")\
**Posts on this page:** 1\
**Page:** 1

<div class="post-metadata">

**Author:** ![CollinsEM](https://yyz2.discourse-cdn.com/flex030/user_avatar/discourse.numenta.org/collinsem/32/160_2.png) [@CollinsEM](https://discourse.numenta.org/u/CollinsEM)\
**Post date:** [August 3, 2024, 12:42pm UTC](https://discourse.numenta.org/t/googles-sparse-autoencoder/11439/1 "2024-08-03T12:42:41Z")

</div>

New research from Google. Apparently they are coming around to sparse representations.

> **[Jumping Ahead: Improving Reconstruction Fidelity with JumpReLU Sparse...](https://arxiv.org/abs/2407.14435)**
>
> Sparse autoencoders (SAEs) are a promising unsupervised approach for identifying causally relevant and interpretable linear features in a language model's (LM) activations. To be useful for downstream tasks, SAEs need to decompose LM activations...