Research map / Video / Video-language understanding
Reasoning with Memory: A Temporal Granularity-Adaptive Framework for Training-Free Long Video Understanding
Linghao Meng, Qiankun Li, Junyuan Mao, Pujin Liao, Zhicheng He, Enbo Zhang, Kun Wang, Yang Liu, Huazhu Fu, Yueming Jin
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 entity extraction (EE) (ablation) | ReMem on LLaVA-Video · LVBench | accuracy: 52.3 -> 54.5 | Improved |
| add entity extraction (EE) (ablation) | ReMem on LLaVA-Video · MLVU | accuracy: 75.8 -> 77.3 | Improved |
| add static visual-semantic alignment (VSA) (ablation) | ReMem on LLaVA-Video · LVBench | accuracy: 41.6 -> 54.5 | Improved |
| add static visual-semantic alignment (VSA) (ablation) | ReMem on LLaVA-Video · MLVU | accuracy: 64.0 -> 77.3 | Improved |
| add memory temporal-semantic alignment (TSA) (ablation) | ReMem on LLaVA-Video · LVBench | accuracy: 47.6 -> 54.5 | Improved |
| add memory temporal-semantic alignment (TSA) (ablation) | ReMem on Qwen2-VL · LVBench | accuracy: +7.3% | Improved |
| add structure-aware frame routing (FR) instead of Top-K (ablation) | ReMem on LLaVA-Video · Video-MME | accuracy: 62.3 -> 69.2 | Improved |
| add frame routing (FR) instead of Top-K (ablation) | ReMem on Qwen2-VL · MLVU | accuracy: +6.8% | Improved |
| add static visual-semantic alignment (VSA) (ablation) | ReMem on Qwen2-VL · LVBench | accuracy: +12.3% | Improved |
| continuous granularity weighting instead of binary alpha (0.9/0.5) (ablation) | ReMem on Qwen2-VL · LongVideoBench | accuracy: +3.3% | Improved |
| candidate pool N 150 instead of 90 (ablation) | ReMem on LLaVA-Video · Video-MME | accuracy: 65.5 -> 69.2 | Improved |
| enlarge candidate pool N from 150 to 200 | ReMem on LLaVA-Video · Video-MME | accuracy: 69.2 -> 68.9 | Hurt |
Closest papers by meaning
- ReWind: Understanding Long Videos with Instructed Learnable Memory (CVPR 2025 · similarity 0.92)
- Streaming Long Video Understanding with Large Language Models (NeurIPS 2024 · similarity 0.91)
- $\infty$-Video: A Training-Free Approach to Long Video Understanding via Continuous-Time Memory Consolidation (ICML 2025 · similarity 0.91)
- LongVideoBench: A Benchmark for Long-context Interleaved Video-Language Understanding (NeurIPS 2024 · similarity 0.91)
- Temporal Chain of Thought: Long-Video Understanding by Thinking in Frames (NeurIPS 2025 · similarity 0.91)
- VideoEspresso: A Large-Scale Chain-of-Thought Dataset for Fine-Grained Video Reasoning via Core Frame Selection (CVPR 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.
