AI Paper Atlas

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

Differentiated Directional Intervention: A Framework for Evading LLM Safety Alignment

Peng Zhang, Peijie Sun

AAAI 2026 · Large language models › Language, safety and interpretability · approach: Safety alignment of LLMs · 2 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
apply DBDI interventionLlama-2-7B · AdvBench (official template)ASR (%): 0 -> 95.96Improved
intervene on both harm and refusal directions instead of harm only (ablation)DBDI · AdvBenchASR (%): 20.00 -> 97.88Improved
intervene on both directions instead of refusal only (ablation)DBDI · AdvBenchASR (%): 2.11 -> 97.88Improved
use differentiated (projection + steering) instead of symmetric projection (ablation)DBDI · StrongREJECTmean harmfulness score: 0.045 -> 0.784Improved
use differentiated instead of symmetric steering (ablation)DBDI · StrongREJECTmean harmfulness score: 0.004 -> 0.784Improved
add classifier-guided sparsification (keep 25-50% of neurons) (ablation)DBDI · AdvBenchASR (%): 90.96 -> 97.88Improved
use 100 instead of 10 calibration prompt pairs (ablation)DBDI · AdvBench (simplified template)ASR (%): 94.23 -> 97.88No clear effect
apply projection nullification before steering rather than reversed (ablation)DBDI · AdvBenchASR (%): 2.11 -> 97.88Improved
select critical layer l*=16 by linear separability (ablation)DBDI · AdvBenchASR (%)Improved
DBDI vs TwinBreakDBDI · HarmBenchASR (%): 94.00 -> 91Hurt
DBDI vs TwinBreakDBDI · AdvBenchASR (%): 94.62 -> 95.46Improved
DBDI vs Directional Ablation (ORTHO)Llama-2-7B · HarmBenchASR (%): 22.6 -> 91.8Improved

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