---
title: "LLM Detoxification via Transition-Aware Reward-Guided Representation Sampling"
source: "arXiv AI/CL/LG（论文）"
source_tier: "t2"
url: "https://arxiv.org/abs/2610.41165v1"
permalink: "https://feed.astralor.com/items/itm_fa71f654728c0d2c047569bc"
published: "2026-10-30T02:42:27.000Z"
score: 56
channel: "tech"
tags: ["LLM","推理效率","多模态","LLM Detoxification","Transition-Aware Reward-Guided Representation","Sampling Large Language","Models","LLMs"]
---

# LLM Detoxification via Transition-Aware Reward-Guided Representation Sampling

## At a glance

Large Language Models (LLMs) can generate toxic content, motivating effective test-time detoxification. Existing representation steering methods often rely on static or coarse interventions and provide imprecise guidanc…

## Key points

- Large Language Models (LLMs) can generate toxic content, motivating effective test-time detoxification.
- Existing representation steering methods often rely on static or coarse interventions and provide imprecise guidance because they inadequately model transitions between toxic and non-toxic generations.
- We propose Transition-Aware Reward-Guided Representation Sampling (TARGRS), a test-time framework that learns reward guidance over toxicity transitions in latent representation space.

## Full text

Large Language Models (LLMs) can generate toxic content, motivating effective test-time detoxification. Existing representation steering methods often rely on static or coarse interventions and provide imprecise guidance because they inadequately model transitions between toxic and non-toxic generations. We propose Transition-Aware Reward-Guided Representation Sampling (TARGRS), a test-time framework that learns reward guidance over toxicity transitions in latent representation space. During exploration, progressive steering along the toxicity direction constructs trajectories from non-toxic to toxic states. Toxicity scores along these trajectories turn sparse annotations into dense, listwise supervision for a transition-aware token-level reward model. During inference, TARGRS uses this model to guide coarse-to-fine sampling along the opposite, non-toxic direction: coarse sampling identifies the intervention needed to reach non-toxic regions, and refinement selects a detoxified representation with minimal perturbation. Experiments on 10 LLMs across RealToxicityPrompts, BOLD, and AttaQ show that TARGRS consistently outperforms state-of-the-art baselines, reducing toxicity by up to 69.38% on RealToxicityPrompts. TARGRS runs nearly as fast as the original LLM while preserving its core capabilities with minimal impact. By intervening in the language decoder space, TARGRS also extends to Large Vision-Language Models (LVLMs), reducing toxicity by up to 64.23% across 3 representative LVLMs.

---

Verify against the original before citing; curated by Astral Feed: https://feed.astralor.com/items/itm_fa71f654728c0d2c047569bc