safety security
Diffusion Decoding Opened a New Bias Control Channel
A feedback controller adjusted activation steering while masked tokens remained revisable.
Summary
A feedback controller adjusted activation steering while masked tokens remained revisable.
On ambiguous bias questions, closed-loop steering raised one model's targeted-group preference from 1.8 to 16.7 percentage points and increased stigmatizing answers on another dataset from 17.6% to 58.1%. The attack assumes internal activation access, but it shows that a diffusion model's repeated denoising steps create an intervention surface absent from one-shot token commitment.
Why it matters
A feedback controller adjusted activation steering while masked tokens remained revisable.
Limits and context
No additional limitation was separately recorded.
Key claims
A feedback controller adjusted activation steering while masked tokens remained revisable.
Evidence: source-2026-10-06-020
Sources
- arXiv preprint 2610.05894arXiv · primary research
Corrections
No corrections have been recorded for this story.