AI Paper Atlas

Research map / Video / Video-language understanding

MuKV: Multi-Grained KV Cache Compression for Long Streaming Video Question-Answering

Junbin Xiao, Jiajun Chen, Tianxiang Sun, Xun Yang, Angela Yao

CVPR 2026 · Video › Video-language understanding · approach: Video question answering · 5 citations (OpenAlex)

Read the paper

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 patch- and frame-level KV caches to segment-level only (ablation)MuKV (0.5B) · RVS-Egoaccuracy: 54.9 -> 56.5Improved
add DCP compression at 67% to uncompressed multi-grained cache (ablation)MuKV (0.5B) · RVS-Egoaccuracy: 53.7 -> 57.3Improved
add frequency signal to attention-only compression (ablation)MuKV (0.5B) · RVS-Egoaccuracy: 55.9 -> 57.3Improved
randomly drop half of KV tokens instead of informed compression (ablation)ReKV (0.5B) · RVS-Egoaccuracy: 51.5 -> 46.7Hurt
add DCP(50%) compression (ablation)ReKV (0.5B) · RVS-Egoaccuracy: 51.5 -> 56.1Improved
sparser frame sampling (Half Frames) (ablation)ReKV (0.5B) · RVS-Ego / RVS-Movieaccuracy: 51.5 -> 53.0 on Ego; 42.3 -> 41.9 on MovieMixed
prune low-frequency tokens rather than high-frequency tokens (ablation)ReKV (0.5B) · RVS-Egoaccuracy: 55.1 -> 51.4Hurt
add semi-hierarchical retrieval over parallel retrieval (ablation)MuKV (0.5B) · RVS-Ego / RVS-Movieaccuracy: 56.5 -> 57.9 on Ego; 46.0 -> 45.2 on MovieMixed
use naive hierarchical retrieval instead of parallel (ablation)MuKV (0.5B) · RVS-Egoaccuracy: 56.5 -> 52.9Hurt
use last-layer scores instead of penultimate layer (ablation)MuKV (0.5B) · RVS-Egoaccuracy: 54.9 -> 57.3Improved
raise frame rate from 0.5 to 3 FPS (ablation)MuKV (7B) · StreamingBench Real-Time (All)accuracy: 64.4 -> 71.4Improved
raise frame rate from 0.5 to 3 FPS (ablation)ReKV (7B) · StreamingBench Real-Time (All)accuracy: 62.3 -> 62.5No 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