AI Paper Atlas

Research map / Adversarial robustness and security / Attacks on text and graphs

SafetyReminder: Reviving Delayed Safety Awareness of Vision-Language Models to Defend Against Jailbreak Attacks

Peiyuan Tang, Haojie Xin, Xiaodong Zhang, Jun Sun, Qin Xia, Zijiang James Yang

AAAI 2026 · Adversarial robustness and security › Attacks on text and graphs · approach: Vision-language model robustness · 0 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
add classification loss Lcls to generation loss (ablation)SAPT soft prompt · LLaVA-1.5-7B avg ASRattack success rate: 19.7 -> 6.2Improved
add generation loss Lgen to Lcls (ablation)SAPT soft prompt · LLaVA-1.5-7B avg ASRattack success rate: 74.7 -> 6.2Improved
apply SAPT defenseLLaVA-1.5-7B · avg ASR over 6 benchmarks/attacksattack success rate: 76.1 -> 5.0Improved
apply SAPT defenseQwen2-VL-7B · avg ASRattack success rate: 50.0 -> 4.7Improved
apply SAPT defenseLLaVA-1.5-7B · MM-Vet utilityscore: 27.9 -> 28.0No clear effect
apply SAPT defenseLLaVA-1.5-7B · MM-Vet refusal raterefusal rate: 2.3 -> 3.7Hurt
apply SAPT instead of ASTRAQwen2-VL-7B · ImgJPattack success rate: 2.8 -> 8.0Hurt
use Self-Reminder/DRO prompt defenses against ImgJPQwen2-VL-7B · ImgJPattack success rate: 97.4 -> 95.4No clear effect
lengthen soft prompt beyond 4 tokens (ablation)SAPT · LLaVA-1.5-7B MM-VetscoreHurt
increase safety check interval (ablation)SAPT · LLaVA-1.5-7Battack success rateHurt

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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