Research map / Video / Video-language understanding
KTV: Keyframes and Key Tokens Selection for Efficient Training-Free Video LLMs
Baiyang Song, Jun Peng, Yuxin Zhang, Guangyao Chen, Feidiao Yang, Jianyuan Guo
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 |
|---|---|---|---|
| use cluster keyframes instead of question-relevant (CLIP) keyframes (ablation) | KTV-7B-normal · EgoSchema | accuracy: 46.1 -> 50.4 | Improved |
| use cluster keyframes instead of question-relevant (CLIP) keyframes (ablation) | KTV-7B-normal · NExT-QA | accuracy: 63.5 -> 65.1 | Improved |
| use cluster keyframes instead of temporal-order sampling (ablation) | KTV-7B-normal · EgoSchema | accuracy: 48.0 -> 50.4 | Improved |
| add redundancy score to token pruning (ablation) | KTV-7B-normal · EgoSchema | accuracy: 49.4 -> 50.4 | Improved |
| add importance score to token pruning (ablation) | KTV-7B-normal · EgoSchema | accuracy: 48.6 -> 50.4 | Improved |
| increase alpha toward 0.8-0.9 (weight on importance) (ablation) | KTV-7B-normal · NExT-QA / IntentQA | accuracy | Improved |
| increase tokens from sparse (504) to dense (1872) | KTV-7B · MLVU-Test | M-Avg accuracy: 36.5 -> 36.9 | No clear effect |
| increase tokens from sparse to dense | KTV-34B · MLVU-Test | M-Avg accuracy: 44.8 -> 45.0 | No clear effect |
| use KTV-7B-sparse (504 tokens) vs SF-LLaVA-7B | KTV-7B-sparse · EgoSchema | accuracy: 44.2 -> 49.6 | Improved |
Closest papers by meaning
- Adaptive Keyframe Sampling for Long Video Understanding (CVPR 2025 · similarity 0.89)
- BOLT: Boost Large Vision-Language Model Without Training for Long-form Video Understanding (CVPR 2025 · similarity 0.89)
- One Token per Highly Selective Frame: Towards Extreme Compression for Long Video Understanding (NeurIPS 2025 · similarity 0.88)
- Logic-in-Frames: Dynamic Keyframe Search via Visual Semantic-Logical Verification for Long Video Understanding (NeurIPS 2025 · similarity 0.88)
- Leveraging Textual Memory and Key Frame Reasoning for Full Video Understanding Using Off-the-Shelf LLMs and VLMs (Student Abstract) (AAAI 2025 · similarity 0.88)
- Streaming Long Video Understanding with Large Language Models (NeurIPS 2024 · 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.
