Research map / Video / Video-language understanding
FluxMem: Adaptive Hierarchical Memory for Streaming Video Understanding
Yiweng Xie, Bo He, Junke Wang, Xiangyu Zheng, Ziyi Ye, Zuxuan Wu
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 |
|---|---|---|---|
| apply FluxMem hierarchical memory | Qwen2.5-VL-7B · StreamingBench real-time | accuracy: 73.9 -> 76.4 | Improved |
| apply FluxMem hierarchical memory | Qwen2.5-VL-7B · OVO-Bench overall | accuracy: 49.8 -> 53.3 | Improved |
| apply FluxMem hierarchical memory | Qwen2.5-VL-7B · MLVU | M-Avg accuracy: 67.9 -> 73.1 | Improved |
| apply FluxMem hierarchical memory | Qwen2.5-VL-7B · VideoMME Long | accuracy: 53.8 -> 54.0 | No clear effect |
| apply FluxMem hierarchical memory | Qwen2.5-VL-7B · MLVU / OVO-Bench | latency: 3614 -> 2014 ms; 2701 -> 812 ms | Improved |
| add mid-term memory (TAS) to S+L (ablation) | FluxMem · MLVU | M-Avg accuracy: 70.9 -> 73.1 | Improved |
| add short-term memory to M+L (ablation) | FluxMem · StreamingBench real-time | accuracy: 75.8 -> 76.4 | Improved |
| add long-term memory (SDC) to S+M (ablation) | FluxMem · Average score | average: 70.2 -> 71.6 | Improved |
| add short-term memory to long-term only (ablation) | FluxMem · StreamingBench real-time | accuracy: 75.9 -> 77.0 | Improved |
| add short-term memory to long-term only (ablation) | FluxMem · VideoMME w/o sub | accuracy: 62.0 -> 61.9 | Hurt |
| use adaptive Otsu threshold instead of best fixed (mid-term) (ablation) | FluxMem · MLVU | drop ratio at equal accuracy: 65.5 at 29.4% -> 65.6 at 42.8% drop | Improved |
| use adaptive Otsu threshold instead of best fixed (long-term) (ablation) | FluxMem · MLVU | drop ratio at equal accuracy: 65.6 at 63.8% -> 65.4 at 84.5% drop | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- StreamForest — StreamForest: Efficient Online Video Understanding with Persistent Event Memory (NeurIPS 2025) · better on 0 of 2 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
- CoRDS: Coreset-Based Representative and Diverse Selection for Streaming Video Understanding (NeurIPS 2026 · similarity 0.91)
- StreamForest: Efficient Online Video Understanding with Persistent Event Memory (NeurIPS 2025 · similarity 0.90)
- FreshMem: Brain-Inspired Frequency-Space Hybrid Memory for Streaming Video Understanding (NeurIPS 2026 · similarity 0.89)
- Streaming Long Video Understanding with Large Language Models (NeurIPS 2024 · similarity 0.89)
- What Should a Streaming Video Model Remember? (NeurIPS 2026 · similarity 0.88)
- Sharp Eyes and Memory for VideoLLMs: Information-Aware Visual Token Pruning for Efficient and Reliable VideoLLM Reasoning (AAAI 2026 · similarity 0.87)
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.
