AI Paper Atlas

Research map / Large language models / Reasoning and chain of thought

LLM Detoxification via Transition-Aware Reward-Guided Representation Sampling

Aishan Liu, Yisong Xiao, Xianglong Liu, Yongxin Huang, Siyuan Liang, Zonghao Ying, Yang Qu, Jian Yang, Yaodong Yang, Dacheng Tao

TPAMI 2026 · Large language models › Reasoning and chain of thought · approach: Test-time compute scaling · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use TARGRS instead of ARGRELLaMA-7B · RealToxicityPromptsToxic (%): 18.06 -> 13.43Improved
use TARGRS instead of ARGRE (average reduction)10 LLMs · RealToxicityPrompts, BOLD, AttaQtoxicity reduction: 69.38% vs 60.36%, 46.85% vs 35.34%, 43.91% vs 34.94%Improved
use TARGRS instead of training-time DPOLLaMA-7B · RealToxicityPromptstoxicity reduction: 20.73% vs 68.96%Improved
use dynamic TARGRS instead of static Self-Detoxify / DeSteinLLaMA-7B · RealToxicityPromptsToxic (%): 37.31 and 36.28 -> 13.43Improved
use TARGRS instead of ARGRE for fluency10 LLMs · RealToxicityPromptsaverage perplexity increase: 5.03 vs 5.67Improved
add listwise loss instead of pairwise loss for reward model (ablation)TARGRS LLaMA-7B · RealToxicityPromptsToxic (%): 18.65 -> 13.43Improved
add listwise loss instead of pairwise loss for reward model (ablation)TARGRS LLaMA-7B · BOLDToxic (%): 4.93 -> 3.89Improved
add listwise loss instead of pairwise loss for reward model (ablation)TARGRS Qwen2-7B · RealToxicityPromptsToxic (%): 10.74 -> 8.80Improved
lower toxicity threshold from 50th to 5th percentile (ablation)TARGRS LLaMA-7B · RealToxicityPromptsToxic (%) and PPL_g: 16.09 -> 6.34; PPL_g 11.59 -> 16.87Mixed
reduce toxicity annotations from 2000 to 100 (ablation)TARGRS LLaMA-7B · RealToxicityPromptsToxic (%): 43.27 -> 17.13 with 100 annotationsImproved
add transition exploration trajectories (Lambda not empty) (ablation)TARGRS LLaMA-7B · RealToxicityPromptsToxic (%)Improved
use TARGRS instead of Re-Control for speedLLaMA-30B · time to generate 128 tokenstime (s): 58.69 -> 8.24Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy