Research map / Video / Video-language understanding
VideoLLaMB: Long Streaming Video Understanding with Recurrent Memory Bridges
Yuxuan Wang, Yiqi Song, Cihang Xie, Yang Liu, Zilong Zheng
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 |
|---|---|---|---|
| replace recurrent memory with mean pooling (ablation) | VideoLLaMB · EgoSchema | accuracy: 51.61 (-2.19) | Improved |
| add recurrent memory instead of adaptive pooling (ablation) | VideoLLaMB · EgoSchema | accuracy: 49.4 (-4.4) | Improved |
| add memory retrieval to the memory cache (ablation) | VideoLLaMB · EgoSchema | accuracy: 52.2 (-1.6) | Improved |
| use SceneTiling instead of uniform segmentation (ablation) | VideoLLaMB · EgoSchema | accuracy: 52 (-1.8) | Improved |
| add image data mixture in training (ablation) | VideoLLaMB · EgoSchema | accuracy: 49.8 (-4.0) | Improved |
| add bridge-layer output rather than memory tokens only (ablation) | VideoLLaMB · EgoSchema | accuracy: 50.4 (-3.4) | Improved |
| use k=8 segments instead of default (ablation) | VideoLLaMB · EgoSchema | accuracy: 52.8 (-1.0) | Hurt |
| increase memory tokens 32 to 64 with 1 bridge layer (ablation) | VideoLLaMB · EgoSchema | accuracy: 53.8 -> 53 | Hurt |
| increase bridge layers 1 to 3 and memory tokens 32 to 64 (ablation) | VideoLLaMB · EgoSchema | accuracy: 53.8 -> 54.6 | Improved |
| VideoLLaMB bridge memory instead of adaptive pooling compression (ablation) | VideoLLaMB · EgoSchema | accuracy: 45.6 -> 53.8 | Improved |
| train on VideoChat2 data instead of PLLaVA data | VideoLLaMB · MVBench | average accuracy: 49.33 -> 52.5 | Improved |
| dynamic segments instead of static 4 segments for length extrapolation (ablation) | VideoLLaMB · EgoSchema | accuracy | Mixed |
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.
- MALMM — MA-LMM: Memory-Augmented Large Multimodal Model for Long-Term Video Understanding (CVPR 2024) · better on 1 of 1 shared result
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
- ReWind: Understanding Long Videos with Instructed Learnable Memory (CVPR 2025 · similarity 0.90)
- Streaming Long Video Understanding with Large Language Models (NeurIPS 2024 · similarity 0.90)
- Unleashing Hour-Scale Video Training for Long Video-Language Understanding (NeurIPS 2025 · similarity 0.90)
- HIMMEL: Hierarchical Interleaved Multi-stream Motion Encoding for Long Video Understanding (NeurIPS 2026 · similarity 0.89)
- FastVID: Dynamic Density Pruning for Fast Video Large Language Models (NeurIPS 2025 · similarity 0.89)
- VideoRefer Suite: Advancing Spatial-Temporal Object Understanding with Video LLM (CVPR 2025 · similarity 0.89)
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.
