---
title: Steering Externalities
emoji: 🧭
colorFrom: blue
colorTo: gray
sdk: static
app_file: index.html
pinned: false
short_description: Benign activation steering and jailbreak risk
---

# Steering Externalities

**Benign Activation Steering Unintentionally Increases Jailbreak Risk for LLMs**

NeurIPS 2026. Chen Xiong and Zhiyuan He contributed equally as co-first authors.

This static research project page includes the paper PDF, original figures, an interactive results explorer, and BibTeX. Results are transcribed from the camera-ready manuscript, with main results and five-resample statistics kept separate.

The Code button currently links to the repository listed in the manuscript.

Page design inspired by [CoP](https://huggingface.co/spaces/TrustSafeAI/CoP) and the [Academic Project Page Template](https://github.com/eliahuhorwitz/Academic-project-page-template). Page code is newly written.