AI Paper Atlas

Research map / Video / Video-language understanding

Question-guided Visual Compression with Memory Feedback for Long-Term Video Understanding

Sosuke Yamao, Natsuki Miyahara, Yuankai Qi, Shun Takeuchi

CVPR 2026 · Video › Video-language understanding · approach: Video LLMs with fine-grained understanding · 2 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
add context memory to vanilla multi-frame compressor (ablation)QViC-MF vanilla · MLVU test / VNBench Longaccuracy: MLVU 48.8 -> 42.4; VNBench Long 39.3 -> 40.9Mixed
add memory feedback (ablation)QViC-MF with context memory · MLVU testaccuracy: 42.4 -> 52.2Improved
add memory feedback (ablation)QViC-MF with context memory · VNBench Longaccuracy: 40.9 -> 55.1Improved
add framewise Ctx2Vis masking (ablation)QViC-MF · VNBench Overall / Longaccuracy: MLVU 52.2 -> 53.0; VNBench Overall 63.7 -> 61.1; Long 55.1 -> 50.7Mixed
add blocking of context-to-text attention (ablation)QMSA · MLVU testaccuracy: 53.0 -> 57.9Improved
add guiding context-to-visual attention (ablation)QMSA · MLVU testaccuracy: 57.9 -> 59.4Improved
add guiding context-to-visual attention (ablation)QMSA · VNBench Longaccuracy: 56.7 -> 58.7Improved
increase sampling from 1 fps to 2 fpsQViC-MF · MLVU testaccuracy: 57.6 -> 59.4Improved
increase sampling from 1 fps to 2 fpsQViC-MF · LVBenchaccuracy: 50.3 -> 50.2No clear effect
use QMSA compressor instead of single-frame or average pooling under high compression (ablation)visual compressor · MLVU / VNBenchaccuracy keep rateImproved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

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