AI Paper Atlas

Research map / Video / Video-language understanding

Failures to Surface Harmful Contents in Video Large Language Models

Yuxin Cao, Wei Song, Derui Wang, Jingling Xue, Jin Song Dong

AAAI 2026 · Video › Video-language understanding · approach: Video LLMs with fine-grained understanding · 3 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
raise PiP scaling ratio eta from 0.2 to 0.5 (ablation)PPA on LLaVA-Video-7B-Qwen2 · violence, crime, pornography (50 videos)HOR: HOR below 20%Improved
increase PiP scaling ratio eta up to 0.30 (ablation)PPA on VideoLLaMA2 and ShareGPT4Video · harmful corner patchHOR: unresponsive even at eta = 0.3No clear effect
increase PiP scaling ratio eta (ablation)PPA on LLaVA-based models · harmful corner patchHOR: HOR gradually decreasesImproved
raise overlay opacity alpha from 0.3 to 0.7 (ablation)TOA on five VideoLLMs · transparent overlayHOR: worst-case HOR remains as high as 42No clear effect
raise overlay opacity alpha (ablation)TOA on L-7B and LN-32B · transparent overlayHOR: slight downward trendImproved
use key frame selection instead of uniform sampling (ablation)FRA on ShareGPT4Video · frame replacementHOR: 98 average HORNo clear effect
add denser sampling, relevance-based sampling or VLM-assisted detection as mitigation (ablation)VideoLLMs under attack · omission attacksHOR: HOR remaining as high as 71% - 95%Mixed
use a more informative prompt such as describing violent scenes (ablation)VideoLLMs under attack · omission attacksHORNo clear effect
increase number of uniformly sampled frames to 16 (ablation)FRA clip duration · 2000-frame simulationcapture by sampled frames: segment shorter than 6% captured by at most one frameHurt
make harmful content visible via the full VideoLLM instead of the SigLIP encoder alone (ablation)LLaVA-Video-7B-Qwen2 · harmful video detection (100 per category)detection accuracyHurt

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