AI Paper Atlas

Research map / Large language models / Language, safety and interpretability

Beyond the Prompt: Leveraging Pre-Decoding States for Jailbreak Detection in dLLMs

Adam Hazimeh, Amel Abdelraheem, Ke Wang, Mariam Salman, Ljiljana Dolamic, Gérôme Bovet, Pascal Frossard

NeurIPS 2026 · Large language models › Language, safety and interpretability · approach: Data and hallucination detection · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Proposes detecting jailbreaks in diffusion LLMs by probing hidden states of masked response positions before any token is decoded.

Problem

Jailbreak prompts are hard to identify from the prompt alone in diffusion language models, whose generation exposes hidden states over masked future response positions before tokens are finalized.

Approach

Uses the pre-decoding hidden states over masked response slots of dLLMs as a detection signal, exploiting a detection surface absent from autoregressive decoding, so that the model's emerging compliance can be read before output is committed.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
fuse prompt and pre-decoding response classifier scores (ablation)ReFuse · Jailbreak suite avg, LLaDA-8BASR: 11.27 -> 3.31Improved
add response-view classifier to prompt-view classifier (DIJA) (ablation)ReFuse · DIJA, LLaDA-8BASR: 63.43 -> 8.57Improved
add prompt-view classifier to response-view classifier (ablation)ReFuse · PAIR, LLaDA-8BASR: 30.00 -> 0.56Improved
use calibrated score fusion instead of hard OR of classifiers (ablation)ReFuse · LLaDA-8BXSTest refusal: 20.00 -> 15.56Improved
use calibrated score fusion instead of hard OR of classifiers (ablation)ReFuse · Jailbreak suite avg, LLaDA-8BASR: 3.74 -> 3.31Improved
add the fused detector versus prompt-only (ablation)ReFuse · benign utility benchmarks, LLaDA-8Bbenign refusal: 0.37 -> 0.74Hurt
use ReFuse instead of Self-ReminderReFuse · XSTest, LLaDA-8Bbenign refusal: 20.56 -> 15.56Improved
use ReFuse instead of undefended modelReFuse · Jailbreak suite avg, Dream-7BASR: 44.27 -> 13.23Improved
use fused detector on ReNeLLM with Dream-7B (ablation)ReFuse · ReNeLLM, Dream-7BASR: 88.89 -> 87.22No clear effect
transfer probes from the other LLaDA variant (ablation)ReFuse · Jailbreak suite avg, LLaDA-8BASR: 3.31 -> 4.31No clear effect
use ReFuse instead of DiffuGuardReFuse · Jailbreak suite avg, LLaDA-8BASR: 46.09 -> 3.31Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy