Research map / Video / Video-language understanding
AdaSpark: Adaptive Sparsity for Efficient Long-Video Understanding
Handong Li, Zikang Liu, Longteng Guo, Tongtian Yue, Yepeng Tang, Xinxin Zhu, Chuanyang Zheng, Ziming Wang, Zhibin Wang, Jun Song, Cheng Yu, Bo Zheng and 1 more
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 AdaS-Attn to the dense model (ablation) | Qwen2.5-VL-3B · TFLOPs | TFLOPs: 299.5 -> 213.1 | Improved |
| add AdaS-FFN to the dense model (ablation) | Qwen2.5-VL-3B · TFLOPs | TFLOPs: 299.5 -> 216.9 | Improved |
| combine AdaS-Attn and AdaS-FFN (ablation) | Qwen2.5-VL-3B · TFLOPs | TFLOPs: 299.5 -> 128.5 | Improved |
| add AdaS-FFN only (ablation) | Qwen2.5-VL-3B · MLVU | M-avg: 65.2 -> 63.9 | Hurt |
| add AdaS-FFN only (ablation) | Qwen2.5-VL-3B · CharadesSTA | mIoU: 43.6 -> 41.7 | Hurt |
| add AdaS-FFN only (ablation) | Qwen2.5-VL-3B · VideoMME | accuracy: 60.3 -> 62.8 | Improved |
| add AdaS-Attn only (ablation) | Qwen2.5-VL-3B · MLVU | M-avg: 65.2 -> 66.0 | Improved |
| add mean compensation to AdaS-FFN (ablation) | AdaSpark 3B · MLVU | M-avg: 56.4 -> 57.5 | Improved |
| add mean compensation to AdaS-FFN (ablation) | AdaSpark 3B · VideoMME | accuracy: 54.1 -> 55.7 | Improved |
| add mean compensation to AdaS-FFN (ablation) | AdaSpark 3B · CharadesSTA | mIoU: 35.0 -> 36.1 | Improved |
| use dynamic Top-P instead of static Top-K (ablation) | AdaSpark 3B · VideoMME | accuracy: 52.6 -> 55.7 | Improved |
| use Top-P instead of uniform sampling (ablation) | AdaSpark 3B · VideoMME | accuracy: 50.2 -> 55.7 | 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.
- FrameFusion — FrameFusion: Combining Similarity and Importance for Video Token Reduction on Large Vision Language Models (ICCV 2025) · better on 3 of 3 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
- LongVU: Spatiotemporal Adaptive Compression for Long Video-Language Understanding (ICML 2025 · similarity 0.89)
- AdaCM^2: On Understanding Extremely Long-Term Video with Adaptive Cross-Modality Memory Reduction (CVPR 2025 · similarity 0.89)
- One Token per Highly Selective Frame: Towards Extreme Compression for Long Video Understanding (NeurIPS 2025 · similarity 0.88)
- Video-XL: Extra-Long Vision Language Model for Hour-Scale Video Understanding (CVPR 2025 · similarity 0.87)
- Streaming Long Video Understanding with Large Language Models (NeurIPS 2024 · similarity 0.87)
- FastVID: Dynamic Density Pruning for Fast Video Large Language Models (NeurIPS 2025 · 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.
