← Tuesday, October 6, 2026Models & trainingPaper

A PI controller steers diffusion language models toward targeted answers

huggingface.co by Sarim Hashmi et al. Editor score 8/10

The attack adjusts activation-steering strength during denoising; on ambiguous BBQ questions, it raises preference for a targeted group from 1.8 to 16.7 percentage points. Engineers working with diffusion LMs should account for their repeated exposure to intermediate predictions.

What

A targeted-bias attack steers a frozen masked diffusion language model toward an answer selected by an adversary.

Method

During denoising, a PI controller tracks target-answer probability and adjusts steering-vector strength; on BBQ, target preference rose from 1.8 to 16.7 percentage points.

Why it matters

Because answers remain revisable across denoising steps, model builders must consider attacks that adapt activation interventions before token commitment.

Read on huggingface.co → Code

Headline, summary, brief and diagram are written by a model from the source's own text; the link, byline and numbers come from the source. Read the original before relying on a claim.