AI Paper Atlas

Research map / Video / Video-language understanding

Free-MoRef: Instantly Multiplexing Context Perception Capabilities of Video-MLLMs within Single Inference

Kuo Wang, Quanlong Zheng, Junlin Xie, Yanhao Zhang, Jinguo Luo, Haonan Lu, Liang Lin, Fan Zhou, Guanbin Li

ICCV 2025 · Video › Video-language understanding · approach: Long video understanding · 1 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 Free-MoRef to 128-frame inputLLaVA-Video-7B · VideoMME overallaccuracy: 64.9 -> 66.3Improved
apply Free-MoRef to 256-frame inputLLaVA-Video-7B · MLVUaccuracy: 67.2 -> 72.5Improved
apply Free-MoRef to 512-frame inputLLaVA-Video-7B · VideoMME overallaccuracy: 60.6 -> 66.9Improved
apply Free-MoRef to 512-frame inputLLaVA-Video-7B · LongVideoBench overallaccuracy: 54.3 -> 59.9Improved
double input frames from 64 to 128 without Free-MoRefLLaVA-Video-7B · MLVUaccuracy: 70.3 -> 70.2No clear effect
quadruple input frames from 64 to 256 without Free-MoRefLLaVA-Video-7B · VideoMME overallaccuracy: 64.3 -> 63.1Hurt
add MoRef attention to partition plus fusion (ablation)Free-MoRef · VideoMME overall, 128 framesaccuracy: 62.0 -> 66.3Improved
add Reference Fusion to partition plus MoRef attention (ablation)Free-MoRef · VideoMME overall, 128 framesaccuracy: 65.8 -> 66.3Improved
add Reference Fusion alone (drop 50% tokens, FastV-like) without partition (ablation)LLaVA-Video-7B · VideoMME overall, 128 framesaccuracy: 64.9 -> 63.9Hurt
add multi-reference partition with full attention instead of MoRef attention (ablation)Reference Fusion baseline · VideoMME overall, 128 framesaccuracy: 63.9 -> 62.0Hurt
increase chunk number N from 2 to 8 (ablation)Free-MoRef · VideoMME overall, 128 framesaccuracy: 66.3 -> 65.9Hurt
change temporal units M from 1 to 64 (ablation)Free-MoRef · VideoMME temporal perceptionaccuracy: 70.9 -> 74.5Improved

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