{"slug": "noise-out-bias-in-targeted-bias-injection-in-diffusion-language-models-via-loop", "title": "Noise Out, Bias In: Targeted Bias Injection in Diffusion Language Models via Closed-Loop Activation Steering", "summary": "Researchers identified a targeted bias injection attack on masked diffusion language models (dLLMs) that exploits their iterative denoising process, in which each token's distribution is re-exposed at every denoising step rather than once at commit time as in autoregressive decoders. The attack, described as closed-loop activation steering, uses this repeated exposure to steer model outputs toward injected biases. The finding matters because dLLMs' multi-step re-prediction of each token gives attackers a wider surface for manipulation than autoregressive models provide.", "body_md": "Masked diffusion language models (dLLMs) generate text by iteratively denoising masked positions, re-predicting each token multiple times before it is committed. An autoregressive decoder exposes an answer's distribution once, at the step that commits it; a dLLM exposes it at every denoising step be", "url": "https://wpnews.pro/news/noise-out-bias-in-targeted-bias-injection-in-diffusion-language-models-via-loop", "canonical_source": "https://aiflash.com/news/131765/", "published_at": "2026-10-06 05:00:13+00:00", "updated_at": "2026-10-06 05:17:40.911449+00:00", "lang": "en", "topics": ["artificial-intelligence", "large-language-models", "ai-safety", "ai-research"], "entities": [], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/noise-out-bias-in-targeted-bias-injection-in-diffusion-language-models-via-loop", "markdown": "https://wpnews.pro/news/noise-out-bias-in-targeted-bias-injection-in-diffusion-language-models-via-loop.md", "text": "https://wpnews.pro/news/noise-out-bias-in-targeted-bias-injection-in-diffusion-language-models-via-loop.txt", "jsonld": "https://wpnews.pro/news/noise-out-bias-in-targeted-bias-injection-in-diffusion-language-models-via-loop.jsonld"}}