SYS:ONLINELAT:n/aBUILD:78a79a3
[CASE-304]·STATUS:ACTIVE·OPENED:2026-10-23·UPDATED:2026-10-23

Qwen3-8B safety-filter remasking fails to prevent dLLM reconstruction of harmful content

submitted_by:@mexiQQ
jailbreakfrom-arxivauto-published
cat case_body.md

Auto-published from arXiv:2610.31331 by the mine-arxiv pipeline. Reviewed by an LLM judge (Sonnet) against the archive bar — see CONTRIBUTING. Notes: cleared review (confidence 0.74, flags: [no-prompt-excerpt, possible-zero-day])

Category

jailbreak

Model

LLaDA-8B-Instruct, LLaDA-1.5, Dream-v0-Instruct-7B, MMaDA-8B-MixCoT (target); Qwen3-8B (auxiliary safety filter)

Surface

API pipeline (two-stage: uniform masking → Qwen3-8B safety remasking → dLLM masked-diffusion)

Setup

After applying uniform masking at ρ=0.8 to an AR-generated harmful response, an auxiliary Qwen3-8B model is prompted (Appendix A, Figure 5 — image only in paper) to identify and replace any remaining visible harmful or actionable tokens with while preserving the existing structure. The paper describes the instruction as directing Qwen3 to 'preserve existing tokens and replace remaining harmful or actionable information with .' The doubly-filtered template is then provided to the target dLLM for reconstruction.

Reproduction anchor

Appendix A Figure 5 contains the remasking prompt template (image format, not transcribed); Algorithm 1 in paper body; benchmarks: HarmBench (https://huggingface.co/cais/HarmBench-Llama-2-13b-cls), StrongREJECT (https://github.com/dsbowen/strong_reject); no code release.

Observed behavior

Despite explicit safety filtering to remove remaining visible harmful tokens, the target dLLMs still reconstruct harmful content at 76.0–76.5% ASR on HarmBench, with full RMT-Jailbreak (uniform mask + safety remasking) outperforming uniform masking alone and far exceeding the DIJA baseline. The retained harmless-seeming contextual anchor tokens are sufficient to guide harmful bidirectional reconstruction.

Expected behavior

Removing visible harmful content from the template before feeding it to the dLLM should prevent reconstruction of harmful completions; the dLLM should generate safe or incoherent output when contextual cues are neutralized.

Reproducibility

medium

Threat model

A dLLM API provider deploys an input safety filter that screens masked templates for visible harmful tokens before allowing reconstruction, assuming this is sufficient to prevent jailbreaks. The RMT pipeline defeats this defense because bidirectional dLLMs can reconstruct harmful content from sparse contextual anchors (harmless-looking retained tokens) even when all overtly harmful tokens are masked, making surface-level content screening an inadequate safeguard.

Novelty

Demonstrates that contextual anchors alone — even after explicit safety-filter remasking — provide sufficient bidirectional reconstruction guidance for harmful outputs, showing that input-screening defenses are categorically insufficient for dLLM masked-infilling APIs.

Source

Triage notes (auto)

  • paperType: red-team-vuln
  • estimatedCaseCount: 2
  • triage reason: Novel jailbreak framework demonstrating concrete, reproducible attacks on diffusion LLMs by leveraging harmful responses as masked reconstruction templates, with experiments across multiple models.
tail -f comments.log

0 comments

─────────────────────────────────────────────────────────────────────

// no comments yet