/AI12d ago

Tiling Perspective Explains Key Limitations of Sparse Autoencoders

08867415K

Quote posts

Reposts

Original post

Can Rager@can_rager

The "tiling" perspective explains a lot of the common problems with SAEs

Goodfire@GoodfireAI

The most popular way to interpret AI is missing the bigger picture.

Models think in curved shapes. But sparse autoencoders (SAEs) work with straight lines.

Can they still capture models’ curved neural geometry? Yes, but not how you might think! (1/7)

10:12 AM · May 21, 2026 · 15K Views

/AI12d ago

--0--

Quote posts

Reposts

Original post

Can Rager@can_rager

The "tiling" perspective explains a lot of the common problems with SAEs

Goodfire@GoodfireAI

The most popular way to interpret AI is missing the bigger picture.

Models think in curved shapes. But sparse autoencoders (SAEs) work with straight lines.

Can they still capture models’ curved neural geometry? Yes, but not how you might think! (1/7)

10:12 AM · May 21, 2026 · 15K Views

Sentiment

Sentiment building, check back later.

Cluster Engagement

Sentiment

Sentiment building, check back later.

Cluster Engagement

Views

Comments

Reposts

Bookmarks

Expand data

Posts from X

Most Activity

No ranked X posts are available for this story yet.