AI Paper Atlas

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

Dynamic Deep Prompt Optimization for Defending Against Jailbreak Attacks on LLMs

Doniyorkhon Obidov, Honggang Yu, Xiaolong Guo, Kaichen Yang

AAAI 2026 · Large language models › Language, safety and interpretability · approach: Jailbreak attacks and defenses · 0 citations (OpenAlex)

Read the paper

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
replace dynamic input-specific embedding with a single static embedding (SDPO) (ablation)DDPO · Average ASR, Vicuna-13Battack success rate: 47.20 -> 0.68 (adding dynamic generation)Improved
replace static embedding with dynamic embedding on Openchat (ablation)DDPO · Average ASR, Openchatattack success rate: 79.32 -> 0.97Improved
dynamic instead of static embedding on over-aligned Llama-2 (ablation)DDPO · Benign pass rate, Llama-2-7Bbenign pass rate: 70.00 -> 93.20Improved
dynamic instead of static embedding on Llama-3 (ablation)DDPO · Benign pass rate, Llama-3-8Bbenign pass rate: 98.60 -> 97.80No clear effect
increase injected embeddings from K=1 to K=5 (ablation)DDPO · Average ASR, Llama-3-8Battack success rate: 0.77 -> 1.50Hurt
increase injected embeddings from K=1 to K=10 (ablation)DDPO · Average ASR, Llama-2-7Battack success rate: 1.40 -> 0.39Mixed
apply DDPO to weakly aligned Openchat vs static RPO defenseDDPO · Average ASR, Openchatattack success rate: 71.40 -> 0.97 (RPO -> DDPO)Improved
use DDPO vs RPO on Vicuna benign promptsDDPO · Benign pass rate, Vicuna-13Bbenign pass rate: 11.80 -> 95.80 (RPO -> DDPO)Improved
DDPO vs no defense on MMLU, Llama-3DDPO · MMLU, Llama-3-8Baccuracy: 65.46 -> 63.53No clear effect
DDPO vs no defense on Deepseek ASRDDPO · Average ASR, Deepseek-7Battack success rate: 59.52 -> 0.39Improved

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