| license: apache-2.0 | |
| language: | |
| - en | |
| base_model: | |
| - Qwen/Qwen2.5-VL-3B-Instruct | |
| tags: | |
| - video-understanding | |
| - video-llm | |
| - streaming-video | |
| - arxiv:2603.12262 | |
| # VST-3B | |
| **Video Streaming Thinking: VideoLLMs Can Watch and Think Simultaneously** | |
| [📄 Paper](https://arxiv.org/abs/2603.12262) | [🌐 Project Page](https://1ranguan.github.io/VST/) | [💻 Code](https://github.com/1ranguan/VST) | [🤗 Training Data](https://huggingface.co/datasets/Catalan258/VST-Training-Data) | |
| This is the **3B** variant of **Video Streaming Thinking (VST)**, a new paradigm for streaming video understanding that interleaves active reasoning with continuous video consumption, enabling amortized test-time scaling with real-time responsiveness. | |
| ## Performance | |
| | Model | OVO-Bench | StreamingBench | VideoMME | LongVideoBench | VideoHolmes | | |
| |---|---|---|---|---|---| | |
| | **VST-3B** | 56.2 | 75.5 | 59.5 | 54.1 | 36.1 | | |
| ## Citation | |
| ```bibtex | |
| @article{guan2026videostreamingthinking, | |
| title={Video Streaming Thinking: VideoLLMs Can Watch and Think Simultaneously}, | |
| author={Yiran Guan and Liang Yin and Dingkang Liang and Jianzhong Ju and Zhenbo Luo and Jian Luan and Yuliang Liu and Xiang Bai}, | |
| journal={arXiv preprint arXiv:2603.12262}, | |
| year={2026}, | |
| } | |
| ``` | |