# Steering and probing for LLM safety

Area: [Mechanistic interpretability](https://mlsanity.com/mechinterp/index.md). HTML: https://mlsanity.com/mechinterp/steering-and-probing-for-llm-safety/

## Key facts

- Steering and probing for LLM safety: 119 papers on arXiv since 2023 (data through 14 Oct 2026).
- 17 of its 119 mapped papers (14%) cite “Programming Refusal with Conditional Activation Steering” (2024, arXiv 2409.05907).
- 14 of its 119 mapped papers (12%) cite “Steering Language Model Refusal with Sparse Autoencoders” (2024, arXiv 2411.11296).
- 10 new papers 24 Sep–14 Oct.

## Most cited here

- [Programming Refusal with Conditional Activation Steering](https://mlsanity.com/paper/2409.05907/) (2024), cited by 17 papers here
- [Steering Language Model Refusal with Sparse Autoencoders](https://mlsanity.com/paper/2411.11296/) (2024), cited by 14 papers here
- [Steering Without Side Effects: Improving Post-Deployment Control of Language Models](https://mlsanity.com/paper/2406.15518/) (2024), cited by 8 papers here
- [The Rogue Scalpel: Activation Steering Compromises LLM Safety](https://mlsanity.com/paper/2509.22067/) (2025), cited by 6 papers here
- [Steering Away from Harm: An Adaptive Approach to Defending Vision Language Model Against Jailbreaks](https://mlsanity.com/paper/2411.16721/) (2024), cited by 5 papers here
- [Understanding Refusal in Language Models with Sparse Autoencoders](https://mlsanity.com/paper/2505.23556/) (2025), cited by 5 papers here
- [Trojan Activation Attack: Red-Teaming Large Language Models using Activation Steering for Safety-Alignment](https://mlsanity.com/paper/2311.09433/) (2023), cited by 4 papers here
- [Revisiting Jailbreaking for Large Language Models: A Representation Engineering Perspective](https://mlsanity.com/paper/2401.06824/) (2024), cited by 4 papers here
- [AlphaSteer: Learning Refusal Steering with Principled Null-Space Constraint](https://mlsanity.com/paper/2506.07022/) (2025), cited by 4 papers here
- [Mechanistic interpretability for steering vision-language-action models](https://mlsanity.com/paper/2509.00328/) (2025), cited by 4 papers here

## New (24 Sep–14 Oct)

- [Nullify: Null-Space Activation Steering for Training-Free LLM Unlearning](https://mlsanity.com/paper/2610.10655/) (2026-10-07)
- [SafeEvo: Deciphering the Safety Alignment Mechanism and Evolution in Language Models](https://mlsanity.com/paper/2610.09600/) (2026-10-07)
- [Backdooring Sparse Autoencoders](https://mlsanity.com/paper/2610.06049/) (2026-10-05)
- [Don't Judge an LLM Only by Its Activations: Discovering Suppressed Safety Features via Counterfactual Activation Potential](https://mlsanity.com/paper/2610.05541/) (2026-10-04)
- [EmoRSS: Mitigating Emotion-Induced Over-Refusal in Large Language Models](https://mlsanity.com/paper/2610.04998/) (2026-10-04)
- [Detect and Suppress: A Mechanistic Defense against Adversarial Patches in VLA Models](https://mlsanity.com/paper/2610.03498/) (2026-10-02)
- [CounterSteer: Suppressing Indirect Prompt Injection with Activation Steering](https://mlsanity.com/paper/2609.36570/) (2026-09-29)
- [Less Sycophancy, Stronger Refusal? Lessons for AI Safety from Mechanistic Interpretability](https://mlsanity.com/paper/2609.35544/) (2026-09-28)
- [When Do Model Internals Help? Exploring the Role of Representation Engineering in LLM Safety](https://mlsanity.com/paper/2609.34771/) (2026-09-28)
- [STR: Supervised Transcoder Replacement for Reducing Steering Side Effects](https://mlsanity.com/paper/2609.32519/) (2026-09-26)