Metraipolis › archive › dc_c1adea363caa
Beyond Token Positions: Safety Alignment Across Denoising Steps in Diffusion Language Models — arXiv:2609.00495
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.00495
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-17T06:33:25.881Z
- SHA-256 of the archived text
0fc745885578ded707f267e55ac23bc28d3e969d298da74100195d245f0c8a3a- Stable URL of this page
https://metraipolis.com/documento/dc_c1adea363caa
How to cite
Beyond Token Positions: Safety Alignment Across Denoising Steps in Diffusion Language Models — arXiv:2609.00495 — arXiv (en), https://arxiv.org/abs/2609.00495 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_c1adea363caa (Metraipolis archive dc_c1adea363caa, sha256 0fc745885578ded707f267e55ac23bc28d3e969d298da74100195d245f0c8a3a)
BibTeX
@misc{dcc1adea363caa,
title = {Beyond Token Positions: Safety Alignment Across Denoising Steps in Diffusion Language Models — arXiv:2609.00495},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.00495},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_c1adea363caa; sha256 0fc745885578ded707f267e55ac23bc28d3e969d298da74100195d245f0c8a3a},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.00495. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Diffusion large language models (dLLMs) generate text through iterative denoising rather than left-to-right decoding. This generation paradigm introduces two axes that can influence safety alignment: when tokens are generated during denoising and where they appear in the response. In this paper, we measure dLLM safety behavior under harmful prompts by tracing intermediate token distributions and commitment decisions throughout denoising. Our analysis shows that refusal signals are concentrated in early denoising steps and leading response positions, and the tokens committed early can strongly shape the final safety outcome. Our measurements further show that the denoising step and persistence of refusal-token commitment are important for understanding dLLM safety. Based on these findings, we propose Refusal-Aware Early Commitment (RAEC), a simple training-free decoding method that commits persistent refusal signals from early steps. Experiments on LLaDA and Dream show that RAEC reduces attack success rates while largely preserving utility. The code is available at this https URL . Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-02 in cs.AI). Categorie arXiv: Computation and Language (cs.CL) ; Artificial Intelligence (cs.AI). Pagina del lavoro: https://arxiv.org/abs/2609.00495 — 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 2026-10-17T06:15:12.789Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-02?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).