# DESTEIN: Navigating Detoxification of Language Models via Universal Steering Pairs and Head-wise Activation Fusion

arXiv: https://arxiv.org/abs/2404.10464 · HTML: https://mlsanity.com/paper/2404.10464/

## Key facts

- Cited by 9 papers in the map (data through 14 Oct 2026).
- Builds on 33 of its 49 references that are in the map.
- First posted on arXiv 16 Apr 2024.

## Builds on (most cited first, 30 of 33)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [GPT-4 Technical Report](https://mlsanity.com/paper/2303.08774/) (2023)
- [Llama 2: Open Foundation and Fine-Tuned Chat Models](https://mlsanity.com/paper/2307.09288/) (2023)
- [Measuring Massive Multitask Language Understanding](https://mlsanity.com/paper/2009.03300/) (2020)
- [OPT: Open Pre-trained Transformer Language Models](https://mlsanity.com/paper/2205.01068/) (2022)
- [Don't Stop Pretraining: Adapt Language Models to Domains and Tasks](https://mlsanity.com/paper/2004.10964/) (2020)
- [Understanding intermediate layers using linear classifier probes](https://mlsanity.com/paper/1610.01644/) (2016)
- [Representation Engineering: A Top-Down Approach to AI Transparency](https://mlsanity.com/paper/2310.01405/) (2023)
- [Inference-Time Intervention: Eliciting Truthful Answers from a Language Model](https://mlsanity.com/paper/2306.03341/) (2023)
- [RealToxicityPrompts: Evaluating Neural Toxic Degeneration in Language Models](https://mlsanity.com/paper/2009.11462/) (2020)
- [Steering Llama 2 via Contrastive Activation Addition](https://mlsanity.com/paper/2312.06681/) (2023)
- [Ethical and social risks of harm from Language Models](https://mlsanity.com/paper/2112.04359/) (2021)
- [Steering Language Models With Activation Engineering](https://mlsanity.com/paper/2308.10248/) (2023)
- [Plug and Play Language Models: A Simple Approach to Controlled Text Generation](https://mlsanity.com/paper/1912.02164/) (2019)
- [The Linear Representation Hypothesis and the Geometry of Large Language Models](https://mlsanity.com/paper/2311.03658/) (2023)
- [Contrastive Decoding: Open-ended Text Generation as Optimization](https://mlsanity.com/paper/2210.15097/) (2022)
- [Knowledge Neurons in Pretrained Transformers](https://mlsanity.com/paper/2104.08696/) (2021)
- [DExperts: Decoding-Time Controlled Text Generation with Experts and Anti-Experts](https://mlsanity.com/paper/2105.03023/) (2021)
- [A Survey on Large Language Model (LLM) Security and Privacy: The Good, the Bad, and the Ugly](https://mlsanity.com/paper/2312.02003/) (2023)
- [GeDi: Generative Discriminator Guided Sequence Generation](https://mlsanity.com/paper/2009.06367/) (2020)
- [FUDGE: Controlled Text Generation With Future Discriminators](https://mlsanity.com/paper/2104.05218/) (2021)
- [Self-Diagnosis and Self-Debiasing: A Proposal for Reducing Corpus-Based Bias in NLP](https://mlsanity.com/paper/2103.00453/) (2021)
- [A Mechanistic Understanding of Alignment Algorithms: A Case Study on DPO and Toxicity](https://mlsanity.com/paper/2401.01967/) (2024)
- [Exploring the Limits of Domain-Adaptive Training for Detoxifying Large-Scale Language Models](https://mlsanity.com/paper/2202.04173/) (2022)
- [Probing LLMs for hate speech detection: strengths and vulnerabilities](https://mlsanity.com/paper/2310.12860/) (2023)
- [Self-Detoxifying Language Models via Toxification Reversal](https://mlsanity.com/paper/2310.09573/) (2023)
- [On the Challenges of Using Black-Box APIs for Toxicity Evaluation in Research](https://mlsanity.com/paper/2304.12397/) (2023)
- [Click: Controllable Text Generation with Sequence Likelihood Contrastive Learning](https://mlsanity.com/paper/2306.03350/) (2023)
- [PREADD: Prefix-Adaptive Decoding for Controlled Text Generation](https://mlsanity.com/paper/2307.03214/) (2023)
- [Spurious Correlations in Reference-Free Evaluation of Text Generation](https://mlsanity.com/paper/2204.09890/) (2022)

## Built on by (most cited first, 9 of 9)

- [MemOS: A Memory OS for AI System](https://mlsanity.com/paper/2507.03724/) (2025)
- [PrivacyRestore: Privacy-Preserving Inference in Large Language Models via Privacy Removal and Restoration](https://mlsanity.com/paper/2406.01394/) (2024)
- [Detoxifying Large Language Models via Autoregressive Reward Guided Representation Editing](https://mlsanity.com/paper/2510.01243/) (2025)
- [IF-GUIDE: Influence Function-Guided Detoxification of LLMs](https://mlsanity.com/paper/2506.01790/) (2025)
- [ToxiFrench: Benchmarking and Enhancing Language Models via CoT Fine-Tuning for French Toxicity Detection](https://mlsanity.com/paper/2508.11281/) (2025)
- [Test-Time Detoxification without Training or Learning Anything](https://mlsanity.com/paper/2602.02498/) (2026)
- [Concept Attractors in LLMs and their Applications](https://mlsanity.com/paper/2601.11575/) (2025)
- [Cleansing the Artificial Mind: A Self-Reflective Detoxification Framework for Large Language Models](https://mlsanity.com/paper/2601.11776/) (2026)
- [Deep Model Merging: The Sister of Neural Network Interpretability -- A Survey](https://mlsanity.com/paper/2410.12927/) (2024)

## Abstract (arXiv)

Despite the remarkable achievements of language models (LMs) across a broad spectrum of tasks, their propensity for generating toxic outputs remains a prevalent concern. Current solutions involving finetuning or auxiliary models usually require extensive computational resources, hindering their practicality in large language models (LLMs). In this paper, we propose DeStein, a novel method that detoxifies LMs by applying representation engineering in activation spaces with lower resource and time costs. Specifically, we derive detoxification vectors from self-induced, universal steering pairs through arithmetic operations in activation spaces. During inference, detoxification is achieved by fusing the detoxification vectors with the original representations in a head-wise manner. Empirical results demonstrate that our method significantly outperforms previous state-of-the-art approaches on various metrics, while also maintaining satisfactory generation quality and diversity. We further validate the practicality and scalability of DeStein with a series of white-box LLMs. The method is open-sourced at https://github.com/LizLizLi/DeStein. Warning: Some example model outputs may contain highly offensive or disturbing text.