Metraipolis › archive › dc_f04e657d855a

LLM Detoxification via Transition-Aware Reward-Guided Representation Sampling — arXiv:2610.41165

Source: arXiv (en)
Original: https://arxiv.org/abs/2610.41165
License (as archived): CC0-1.0
Archived: 2026-11-02T07:28:54.345Z
SHA-256 of the archived text: 7504fb77081106aa48a7378c22647958f67f62c97eb4213b5c5a69cef2956604
Stable URL of this page: https://metraipolis.com/documento/dc_f04e657d855a

How to cite

LLM Detoxification via Transition-Aware Reward-Guided Representation Sampling — arXiv:2610.41165 — arXiv (en), https://arxiv.org/abs/2610.41165 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_f04e657d855a (Metraipolis archive dc_f04e657d855a, sha256 7504fb77081106aa48a7378c22647958f67f62c97eb4213b5c5a69cef2956604)

BibTeX

@misc{dcf04e657d855a,
  title = {LLM Detoxification via Transition-Aware Reward-Guided Representation Sampling — arXiv:2610.41165},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.41165},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_f04e657d855a; sha256 7504fb77081106aa48a7378c22647958f67f62c97eb4213b5c5a69cef2956604},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.41165. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

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.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno).
Categorie arXiv: Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2610.41165 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-11-02T07:26:57.681Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).