---
title: "[Filler] Sparse Autoencoder Feature Geometry in Small Transformers"
description: "How SAE features arrange themselves across layers, and what that implies for circuit-level interpretability."
pubDate: 2026-04-11
tags: ["mech-interp", "SAEs"]
alsoOn:
  - kind: arxiv
    url: "https://arxiv.org/abs/0000.00000"
  - kind: pdf
    url: "/papers/sae-feature-geometry.pdf"
  - kind: scholar
    url: "https://scholar.google.com/citations?user=GNC3dEkAAAAJ&hl=en"
  - kind: lesswrong
    url: "https://lesswrong.com/users/kaushikreddyxyz"
---

_[Filler research summary — replace with the real abstract and write-up.]_

## Abstract

We study the geometric arrangement of sparse autoencoder features across the
residual stream of small transformers, and find _[filler result]_.

## Key figures

_[Filler — figures and discussion go here. This page demonstrates the
research-post template: metadata, tags, and mirror icons below the post.]_