Research map / Video / Video-language understanding
Accelerating Streaming Video Large Language Models via Hierarchical Token Compression
Yiyu Wang, Xuyang Liu, Xiyan Gui, Xinying Lin, Boxue Yang, Chenfei Liao, Tailai Chen, Linfeng Zhang
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 STC-Pruner (75% prune) instead of ToMe/VisionZip | ReKV · OVO-Bench overall | accuracy: 46.4 -> 50.6 | Improved |
| add STC-Pruner vs VidCom2 | ReKV · StreamingBench overall | accuracy: 63.6 -> 63.7 | No clear effect |
| add STC-Cacher & Pruner jointly | ReKV · OVO-Bench overall | accuracy: 52.6 -> 52.0 | Hurt |
| add STC-Cacher & Pruner jointly (30% tokens) | ReKV · StreamingBench overall | accuracy: 69.1 -> 65.2 | Hurt |
| add STC-Cacher & Pruner jointly | ReKV · OVO-Bench | ViT encoding latency (s): 103.7 -> 78.3 | Improved |
| add STC-Cacher & Pruner jointly | ReKV · OVO-Bench | LLM prefilling latency (s): 482.4 -> 263.7 | Improved |
| add STC-Cacher to ViT of streaming models | StreamForest · OVO-Bench overall | accuracy: 54.3 -> 52.3 | Hurt |
| add STC-Cacher | StreamForest · StreamingBench overall | accuracy: 77.3 -> 76.9 | No clear effect |
| add STC-Cacher | LiveCC · OVO-Bench overall | accuracy: 59.7 -> 57.3 | Hurt |
| reuse only attention features in STC-Cacher (ablation) | STC-Cacher · EgoSchema | accuracy: 59.0 -> 26.2 | Hurt |
| reuse only MLP features in STC-Cacher (ablation) | STC-Cacher · EgoSchema | accuracy: 59.0 -> 57.1 | Hurt |
| use both spatial and temporal anchors in STC-Pruner (ablation) | STC-Pruner · OVO-Bench STU | accuracy: 47.2 -> 48.9 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 2 on most shared results.
- StreamForest — StreamForest: Efficient Online Video Understanding with Persistent Event Memory (NeurIPS 2025) · better on 1 of 2 shared results
- Dispider — Dispider: Enabling Video LLMs with Active Real-Time Interaction via Disentangled Perception, Decision, and Reaction (CVPR 2025) · 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
- DyCoke: Dynamic Compression of Tokens for Fast Video Large Language Models (CVPR 2025 · similarity 0.91)
- One Token per Highly Selective Frame: Towards Extreme Compression for Long Video Understanding (NeurIPS 2025 · similarity 0.90)
- PVC: Progressive Visual Token Compression for Unified Image and Video Processing in Large Vision-Language Models (CVPR 2025 · similarity 0.89)
- Less Is More, but Where? Dynamic Token Compression via LLM-Guided Keyframe Prior (NeurIPS 2025 · similarity 0.88)
- Streaming Long Video Understanding with Large Language Models (NeurIPS 2024 · similarity 0.88)
- VL-Cache: Sparsity and Modality-Aware KV Cache Compression for Vision-Language Model Inference Acceleration (ICLR 2025 · similarity 0.88)
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.
