Research map / Video / Video-language understanding
ST-SimDiff: Balancing Spatiotemporal Similarity and Difference for Efficient Video Understanding with MLLMs
Bingjun Luo, Tony Wang, Chaoqi Chen, Xinpeng Ding
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 spatial similarity selection to baseline (ablation) | ST-SimDiff (LLaVA-Video-7B, retain 30%) · VideoMME overall | accuracy (%): 60.3 -> 61.5 | Improved |
| add temporal similarity selection to baseline (ablation) | ST-SimDiff (LLaVA-Video-7B, retain 30%) · VideoMME overall | accuracy (%): 60.3 -> 61.7 | Improved |
| use joint spatio-temporal similarity instead of either alone (ablation) | ST-SimDiff (LLaVA-Video-7B, retain 30%) · VideoMME overall | accuracy (%): 61.7 -> 62.6 | Improved |
| add difference-based event token selection (Diff) (ablation) | ST-SimDiff (LLaVA-Video-7B, retain 30%) · VideoMME overall | accuracy (%): 62.6 -> 63.2 | Improved |
| add difference-based event token selection (Diff) (ablation) | ST-SimDiff (LLaVA-Video-7B, retain 30%) · EgoSchema | accuracy (%): 55.3 -> 56.0 | Improved |
| add difference-based event token selection (Diff) (ablation) | ST-SimDiff (LLaVA-Video-7B, retain 50%) · VideoMME overall | accuracy (%): 63.7 -> 63.8 | Improved |
| add difference-based event token selection (Diff) (ablation) | ST-SimDiff (NVILA-Video-8B, retain 30%) · VideoMME overall | accuracy (%): 59.9 -> 60.2 | Improved |
| add difference-based event token selection (Diff) (ablation) | ST-SimDiff (NVILA-Video-8B, retain 50%) · LongVideoBench overall | accuracy (%): 56.0 -> 56.5 | Improved |
| add similarity modules to baseline (ablation) | ST-SimDiff (NVILA-Video-8B, retain 30%) · VideoMME overall | accuracy (%): 59.3 -> 59.9 | Improved |
| increase token retain ratio from 30% to 50% | ST-SimDiff (LLaVA-Video-7B) · VideoMME overall | accuracy (%): 63.2 -> 63.8 | Improved |
| increase input frames from 64 to 128 | ST-SimDiff (Qwen2.5-VL-7B, retain 50%) · VideoMME overall | accuracy (%): 62.9 -> 66.3 | Improved |
| compress with ST-SimDiff at 30% retention instead of keeping all tokens | LLaVA-Video-7B · VideoMME overall | accuracy (%): 63.3 -> 63.2 | No clear effect |
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 9 of 9 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
- Weakly Supervised Temporal Action Localization via Dual-Prior Collaborative Learning Guided by Multimodal Large Language Models (CVPR 2025 · similarity 0.88)
- DELTAVID: Enhancing Fine-Grained Spatiotemporal Perception with Cross-Video Differences (NeurIPS 2026 · similarity 0.88)
- Adaptive Keyframe Sampling for Long Video Understanding (CVPR 2025 · similarity 0.88)
- MotionBench: Benchmarking and Improving Fine-grained Video Motion Understanding for Vision Language Models (CVPR 2025 · similarity 0.88)
- LongVU: Spatiotemporal Adaptive Compression for Long Video-Language Understanding (ICML 2025 · similarity 0.88)
- Needle In A Video Haystack: A Scalable Synthetic Evaluator for Video MLLMs (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.
