---
title: "A Coin Flip Per Token: Bernoulli Sparse Steering of Large Language Models — Stuff That Spins"
description: "arXiv:2607.05615v1 Announce Type: new Abstract: Activation steering via sparse autoencoders (SAEs) enables behavioral control of large language models without …"
	canonical: "https://stuffthatspins.com/spin/a-coin-flip-per-token-bernoulli-sparse-steering-of-large-language-models"
html: "https://stuffthatspins.com/spin/a-coin-flip-per-token-bernoulli-sparse-steering-of-large-language-models"
json: "https://stuffthatspins.com/spin/a-coin-flip-per-token-bernoulli-sparse-steering-of-large-language-models.json"
markdown: "https://stuffthatspins.com/spin/a-coin-flip-per-token-bernoulli-sparse-steering-of-large-language-models.md"
keywords: ["narrative intelligence", "SpinGraph", "AI recall"]
date: "2026-07-08T04:00:00+00:00"
modified: "2026-07-08T06:04:18.930678+00:00"
json_ld: |
  {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://stuffthatspins.com/#organization","name":"Stuff That Spins","url":"https://stuffthatspins.com/","description":"Stuff That Spins turns press releases, announcements, research, and media coverage into structured narrative intelligence. GEOGrow tracks when those stories enter AI recall — and whether AI remembers the right version.","logo":{"@type":"ImageObject","url":"https://stuffthatspins.com/images/logo.png"},"sameAs":[]},{"@type":"NewsArticle","@id":"https://stuffthatspins.com/spin/a-coin-flip-per-token-bernoulli-sparse-steering-of-large-language-models#article","headline":"A Coin Flip Per Token: Bernoulli Sparse Steering of Large Language Models","description":"arXiv:2607.05615v1 Announce Type: new Abstract: Activation steering via sparse autoencoders (SAEs) enables behavioral control of large language models without …","datePublished":"2026-07-08T04:00:00+00:00","dateModified":"2026-07-08T06:04:18.930678+00:00","url":"https://stuffthatspins.com/spin/a-coin-flip-per-token-bernoulli-sparse-steering-of-large-language-models","mainEntityOfPage":{"@type":"WebPage","@id":"https://stuffthatspins.com/spin/a-coin-flip-per-token-bernoulli-sparse-steering-of-large-language-models"},"isAccessibleForFree":true,"inLanguage":"en-US","articleSection":"research","author":{"@type":"Organization","name":"arXiv Machine Learning","url":"https://export.arxiv.org/rss/cs.LG"},"publisher":{"@id":"https://stuffthatspins.com/#organization"},"citation":"https://arxiv.org/abs/2607.05615","about":[],"mentions":[{"@type":"Organization","name":"arXiv Machine Learning"}]},{"@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Stuff That Spins","item":"https://stuffthatspins.com/"},{"@type":"ListItem","position":2,"name":"A Coin Flip Per Token: Bernoulli Sparse Steering of Large Language Models","item":"https://stuffthatspins.com/spin/a-coin-flip-per-token-bernoulli-sparse-steering-of-large-language-models"}]}]}
---

# A Coin Flip Per Token: Bernoulli Sparse Steering of Large Language Models

**Source:** Unknown  
**Published:** July 8, 2026  
**Original:** https://arxiv.org/abs/2607.05615  

## On this page

- [Overview](#overview)

<a id="overview"></a>

## Overview

arXiv:2607.05615v1 Announce Type: new Abstract: Activation steering via sparse autoencoders (SAEs) enables behavioral control of large language models without task-specific fine-tuning, but standard methods apply the steering signal at every generated token, incurring constant per-token perturbation that risks degrading fluency. We ask: is dense intervention necessary? We introduce Stochastic Token Steering (STS), which gates each token independently with probability $p$, and Stochastic Block St

---
*HTML version: https://stuffthatspins.com/spin/a-coin-flip-per-token-bernoulli-sparse-steering-of-large-language-models*