Cleansing the Artificial Mind: A Self-Reflective Detoxification Framework for Large Language Models

TL;DR

This paper presents a fully self-reflective detoxification framework for large language models that leverages their inherent abilities to detect and correct toxic content without external modules or human intervention, improving safety and coherence.

Contribution

It introduces a novel self-reflective detoxification method utilizing internal mechanisms of LLMs, eliminating reliance on external tools or data annotation for safer text generation.

Findings

Outperforms state-of-the-art detoxification methods on benchmark datasets.

Preserves semantic fidelity while reducing toxicity effectively.

Demonstrates intrinsic self-detoxification capabilities of LLMs.

Abstract

Recent breakthroughs in Large Language Models (LLMs) have revealed remarkable generative capabilities and emerging self-regulatory mechanisms, including self-correction and self-rewarding. However, current detoxification techniques rarely exploit these built-in abilities; instead, they rely on external modules, labor-intensive data annotation, or human intervention --factors that hinder scalability and consistency. In this paper, we introduce a fully self-reflective detoxification framework that harnesses the inherent capacities of LLMs to detect, correct toxic content, and refine LLMs without external modules and data annotation. Specifically, we propose a Toxic Signal Detector --an internal self-identification mechanism, coupled with a systematic intervention process to transform toxic text into its non-toxic counterpart. This iterative procedure yields a contrastive detoxification dataset used to fine-tune the model, enhancing its ability for safe and coherent text generation. Experiments on benchmark datasets such as DetoxLLM and ParaDetox show that our method achieves better detoxification performance than state-of-the-art methods while preserving semantic fidelity. By obviating the need for human intervention or external components, this paper reveals the intrinsic self-detoxification ability of LLMs, offering a consistent and effective approach for mitigating harmful content generation. Ultimately, our findings underscore the potential for truly self-regulated language models, paving the way for more responsible and ethically guided text generation systems.