AI Paper Atlas

Research map / Video / Video-language understanding

SegMo: Co-Designing Content-Aware Sparsity and Locally-Cohesive Segment Parallelism for Efficient VLM Inference

Haojuan Li, Ruohan Tang, Dongzhou Cheng, Zongpu Zhang, Jian Li, Jiaqi Wang

CVPR 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
replace uniform frame sampling with Content-Aware SparsificationQwen2-VL-7B-Instruct (32 frames) · LVBenchaccuracy: 32.44 -> 44.44Improved
replace uniform frame sampling with Content-Aware SparsificationMiniCPM-o 2.6 (32 frames) · LVBenchaccuracy: 34.92 -> 45.84Improved
add Locally-Cohesive Segment Parallelism on top of CAS (ablation)SegMo (Qwen2-VL-7B-Instruct, 32 frames) · LVBenchaccuracy: 44.44 -> 40.75Hurt
add CAS + LSP versus 2-GPU data-parallel baselineQwen2-VL-7B-Instruct · LongVideoBench (32 frames)TTFT (s): 2.490 -> 0.701Improved
add CAS + LSP versus 2-GPU data-parallel baselineQwen2-VL-7B-Instruct · LVBench (64 frames)TTFT (s): 9.460 -> 2.762Improved
add LSP on top of uniform baseline at 64 frames on Video-MME long videosMiniCPM-o 2.6 (64 frames) · Video-MME long (30-60m)accuracy: 59.00 -> 59.00No clear effect
balanced weight w=0.5 instead of w=0.0 (redundancy only) (ablation)CAS · LongVideoBench (64 frames)accuracy: 49.65 -> 51.83Improved
balanced weight w=0.5 instead of w=1.0 (relevance only) (ablation)CAS · Video-MME (128 frames)accuracy: 66.67 -> 69.00Improved
relevance only (w=1) instead of redundancy only (w=0) (ablation)CAS · LongVideoBench vs Video-MMEaccuracy: 49.65 -> 50.58 (LongVideoBench); 68.05 -> 66.67 (Video-MME)Mixed
add Global Context Injection (ablation)SegMo · LongVideoBench overallaccuracy: 48.46 -> 49.83Improved
add Global Context Injection (ablation)SegMo · LongVideoBench T2A tasksaccuracy: 41.07 -> 51.78Improved

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