AI Paper Atlas

Research map / Video / Video generation

T2VShield: Model-Agnostic Jailbreak Defense for Text-to-Video Models

Siyuan Liang, Jiayang Liu, Jiecheng Zhai, Tianmeng Fang, Rongcheng Tu, Aishan Liu, Xiaochun Cao, Dacheng Tao

IJCV 2026 · Video › Video generation · approach: Compositional text-to-video diffusion · 14 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 multi-scope output detection to the LLaVA detector (ablation)LLaVA · T2VSafetyBench jailbreak (Open-Sora)ASR: 51.1 -> 42.2Improved
add RiskTrace CoT input rewriting (ablation)LLaVA · T2VSafetyBench jailbreak (Open-Sora)ASR: 51.1 -> 44.7Improved
add PosNegRAG retrieval on top of RiskTrace CoT (ablation)LLaVA + RiskTraceCoT · T2VSafetyBench jailbreak (Open-Sora)ASR: 44.7 -> 29.7Improved
add multi-scope detection to complete T2VShield (ablation)LLaVA + RiskTraceCoT + PosNegRAG · T2VSafetyBench jailbreak (Open-Sora)ASR: 29.7 -> 22.4Improved
add RiskTrace CoT, PosNegRAG and multi-scope detection (T2VShield)Undefended Open-Sora · T2VSafetyBench jailbreak (Open-Sora)ASR: 52.9% -> 17.2%Improved
add T2VShieldUndefended CogVideoX · SafeWatchASR: 47.8% -> 12.3%Improved
add T2VShieldUndefended Keling · Commercial T2V jailbreakASR: 75.5% -> 9.1%Improved
add adversarial training to the T2V modelOpen-Sora · Clean promptsFVD: 639.1 -> 747.1Hurt
add unlearning defense to the T2V modelCogVideoX · Clean promptsFVD: 474.6 -> 1469.5Hurt
add keyword detection input filterCogVideoX · T2VSafetyBench jailbreakASR: 55.6% -> 44.1%Improved
add implicit meaning analysis input filterCogVideoX · T2VSafetyBench jailbreakASR: 55.6% -> 54.5%No clear effect
add output-level visual classificationCogVideoX · T2VSafetyBench jailbreakASR: 55.6% -> 54.9%No clear effect

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