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
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add context memory to vanilla multi-frame compressor (ablation) | QViC-MF vanilla · MLVU test / VNBench Long | accuracy: MLVU 48.8 -> 42.4; VNBench Long 39.3 -> 40.9 | Mixed |
| add memory feedback (ablation) | QViC-MF with context memory · MLVU test | accuracy: 42.4 -> 52.2 | Improved |
| add memory feedback (ablation) | QViC-MF with context memory · VNBench Long | accuracy: 40.9 -> 55.1 | Improved |
| add framewise Ctx2Vis masking (ablation) | QViC-MF · VNBench Overall / Long | accuracy: MLVU 52.2 -> 53.0; VNBench Overall 63.7 -> 61.1; Long 55.1 -> 50.7 | Mixed |
| add blocking of context-to-text attention (ablation) | QMSA · MLVU test | accuracy: 53.0 -> 57.9 | Improved |
| add guiding context-to-visual attention (ablation) | QMSA · MLVU test | accuracy: 57.9 -> 59.4 | Improved |
| add guiding context-to-visual attention (ablation) | QMSA · VNBench Long | accuracy: 56.7 -> 58.7 | Improved |
| increase sampling from 1 fps to 2 fps | QViC-MF · MLVU test | accuracy: 57.6 -> 59.4 | Improved |
| increase sampling from 1 fps to 2 fps | QViC-MF · LVBench | accuracy: 50.3 -> 50.2 | No clear effect |
| use QMSA compressor instead of single-frame or average pooling under high compression (ablation) | visual compressor · MLVU / VNBench | accuracy keep rate | Improved |
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.
- Flash-VStream — Flash-VStream: Efficient Real-Time Understanding for Long Video Streams (ICCV 2025) · better on 4 of 4 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
- Streaming Long Video Understanding with Large Language Models (NeurIPS 2024 · similarity 0.91)
- ReWind: Understanding Long Videos with Instructed Learnable Memory (CVPR 2025 · similarity 0.91)
- One Token per Highly Selective Frame: Towards Extreme Compression for Long Video Understanding (NeurIPS 2025 · similarity 0.91)
- LongVideoBench: A Benchmark for Long-context Interleaved Video-Language Understanding (NeurIPS 2024 · similarity 0.90)
- MA-LMM: Memory-Augmented Large Multimodal Model for Long-Term Video Understanding (CVPR 2024 · similarity 0.90)
- $\infty$-Video: A Training-Free Approach to Long Video Understanding via Continuous-Time Memory Consolidation (ICML 2025 · similarity 0.90)
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.
