Gespeichert in:
| Hauptverfasser: | , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2026
|
| Schlagworte: | |
| Online-Zugang: | https://arxiv.org/abs/2603.12938 |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866908884187217920 |
|---|---|
| author | Liu, Zikang Guo, Longteng Li, Handong Zhen, Ru He, Xingjian Ji, Ruyi Ren, Xiaoming Zhang, Yanhao Lu, Haonan Liu, Jing |
| author_facet | Liu, Zikang Guo, Longteng Li, Handong Zhen, Ru He, Xingjian Ji, Ruyi Ren, Xiaoming Zhang, Yanhao Lu, Haonan Liu, Jing |
| contents | Real-time understanding of continuous video streams is essential for interactive assistants and multimodal agents operating in dynamic environments. However, most existing video reasoning approaches follow a batch paradigm that defers reasoning until the full video context is observed, resulting in high latency and growing computational cost that are incompatible with streaming scenarios. In this paper, we introduce ThinkStream, a framework for streaming video reasoning based on a Watch--Think--Speak paradigm that enables models to incrementally update their understanding as new video observations arrive. At each step, the model performs a short reasoning update and decides whether sufficient evidence has accumulated to produce a response. To support long-horizon streaming, we propose Reasoning-Compressed Streaming Memory (RCSM), which treats intermediate reasoning traces as compact semantic memory that replaces outdated visual tokens while preserving essential context. We further train the model using a Streaming Reinforcement Learning with Verifiable Rewards scheme that aligns incremental reasoning and response timing with the requirements of streaming interaction. Experiments on multiple streaming video benchmarks show that ThinkStream significantly outperforms existing online video models while maintaining low latency and memory usage. Code, models and data will be released at https://github.com/johncaged/ThinkStream |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_12938 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Thinking in Streaming Video Liu, Zikang Guo, Longteng Li, Handong Zhen, Ru He, Xingjian Ji, Ruyi Ren, Xiaoming Zhang, Yanhao Lu, Haonan Liu, Jing Computer Vision and Pattern Recognition Artificial Intelligence Real-time understanding of continuous video streams is essential for interactive assistants and multimodal agents operating in dynamic environments. However, most existing video reasoning approaches follow a batch paradigm that defers reasoning until the full video context is observed, resulting in high latency and growing computational cost that are incompatible with streaming scenarios. In this paper, we introduce ThinkStream, a framework for streaming video reasoning based on a Watch--Think--Speak paradigm that enables models to incrementally update their understanding as new video observations arrive. At each step, the model performs a short reasoning update and decides whether sufficient evidence has accumulated to produce a response. To support long-horizon streaming, we propose Reasoning-Compressed Streaming Memory (RCSM), which treats intermediate reasoning traces as compact semantic memory that replaces outdated visual tokens while preserving essential context. We further train the model using a Streaming Reinforcement Learning with Verifiable Rewards scheme that aligns incremental reasoning and response timing with the requirements of streaming interaction. Experiments on multiple streaming video benchmarks show that ThinkStream significantly outperforms existing online video models while maintaining low latency and memory usage. Code, models and data will be released at https://github.com/johncaged/ThinkStream |
| title | Thinking in Streaming Video |
| topic | Computer Vision and Pattern Recognition Artificial Intelligence |
| url | https://arxiv.org/abs/2603.12938 |