AI Paper Atlas

Research map / Video / Video-language understanding

Think in Cloud, Look at Edges: Semantic-Driven Query Decomposition for Efficient Video Reasoning

Wenhao Zou, Zhijie Cai, Minchen Yu, Zongshuai Zhang, Guangxu Zhu

ICML 2026 · Video › Video-language understanding · approach: Long video understanding

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 cloud query decomposition (vs flat Top-K) (ablation)SCOPE 3B · Video-MME K=16accuracy: 55.04 -> 56.08Improved
add semantic-driven budget allocation (vs random) (ablation)SCOPE 3B · Video-MME K=16accuracy: 56.08 -> 56.32Improved
add semantic-driven budget allocation (vs random) (ablation)SCOPE 3B · Video-MME K=32accuracy: 57.17 -> 59.33Improved
add semantic-driven budget allocation (vs random) (ablation)SCOPE 3B · Video-MME K=64accuracy: 59.78 -> 60.07No clear effect
add query decomposition (w/ Decomp. vs w/o) (ablation)Top-K selection, Qwen2.5-VL-72B · Video-MME K=16accuracy: 62.93 -> 66.04Improved
add query decomposition (w/ Decomp. vs w/o) (ablation)Top-K selection, Qwen2.5-VL-72B · Video-MME K=64accuracy: 68.56 -> 70.03Improved
use SCOPE instead of AKSQwen2.5-VL-72B · Video-MME K=16accuracy: 64.2 -> 66.0Improved
use SCOPE instead of AKSQwen2.5-VL-32B · Video-MME K=16accuracy: 60.5 -> 63.0Improved
use SCOPE instead of Uniform samplingQwen2.5-VL-72B · Video-MME K=16accuracy: 62.6 -> 66.0Improved
use SCOPE instead of Top-KQwen2.5-VL-32B · LongVideoBench K=64accuracy: 62.9 -> 64.0Improved
use SCOPE instead of Top-K (edge latency)Qwen2.5-VL-72B · Video-MME K=16latency (s): 14.26 -> 15.26Hurt
use edge-cloud SCOPE instead of pure cloudQwen2.5-VL-72B · Video-MME K=16latency (s): 154.22 -> 23.94Improved

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