X-Stream: Exploring MLLMs as Multiplexers for Multi-Stream Understanding
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , , , , , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2026
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866918535326859264 |
|---|---|
| author | Sun, Peiwen Lu, Xudong Liu, Huadai Bo, Yang Wu, Dongming Guan, Huankang Cai, Minghong Chen, Jinpeng Guo, Xintong Li, Shuhan Liu, Rui Yue, Xiangyu |
| author_facet | Sun, Peiwen Lu, Xudong Liu, Huadai Bo, Yang Wu, Dongming Guan, Huankang Cai, Minghong Chen, Jinpeng Guo, Xintong Li, Shuhan Liu, Rui Yue, Xiangyu |
| contents | While video streaming understanding has made significant strides, real-world applications, such as live sports broadcasting, autonomous driving, and multi-screen collaboration, inherently demand continuous, multi-stream interactions. However, existing benchmarks are confined to single-stream paradigms, leaving a critical gap in evaluating online, cross-stream reasoning. To bridge this, we introduce X-Stream, the first benchmark dedicated to multi-stream streaming understanding. Comprising 4,220 rigorously curated QA pairs across 932 videos, X-Stream evaluates 11 subtasks across multi-window, multi-view, and multi-device scenarios. Crucially, our dataset is constructed using a novel dual-verification pipeline that prevents over-reliance on a single stream. Furthermore, we pioneer the conceptualization of multi-modal large language models (MLLMs) as naive multiplexers, systematically evaluating their performance through the lens of Signal Multiplexing Theory. Our extensive online inference experiments reveal a stark reality: state-of-the-art MLLMs struggle significantly with concurrent streams, achieving only about 50% score and exhibiting poor proactive ability. Ultimately, X-Stream exposes the trade-off of current multiplexing schemes, providing both a practical evaluation protocol and empirical guidance for next-generation multi-stream agents. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2606_02482 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | X-Stream: Exploring MLLMs as Multiplexers for Multi-Stream Understanding Sun, Peiwen Lu, Xudong Liu, Huadai Bo, Yang Wu, Dongming Guan, Huankang Cai, Minghong Chen, Jinpeng Guo, Xintong Li, Shuhan Liu, Rui Yue, Xiangyu Computer Vision and Pattern Recognition While video streaming understanding has made significant strides, real-world applications, such as live sports broadcasting, autonomous driving, and multi-screen collaboration, inherently demand continuous, multi-stream interactions. However, existing benchmarks are confined to single-stream paradigms, leaving a critical gap in evaluating online, cross-stream reasoning. To bridge this, we introduce X-Stream, the first benchmark dedicated to multi-stream streaming understanding. Comprising 4,220 rigorously curated QA pairs across 932 videos, X-Stream evaluates 11 subtasks across multi-window, multi-view, and multi-device scenarios. Crucially, our dataset is constructed using a novel dual-verification pipeline that prevents over-reliance on a single stream. Furthermore, we pioneer the conceptualization of multi-modal large language models (MLLMs) as naive multiplexers, systematically evaluating their performance through the lens of Signal Multiplexing Theory. Our extensive online inference experiments reveal a stark reality: state-of-the-art MLLMs struggle significantly with concurrent streams, achieving only about 50% score and exhibiting poor proactive ability. Ultimately, X-Stream exposes the trade-off of current multiplexing schemes, providing both a practical evaluation protocol and empirical guidance for next-generation multi-stream agents. |
| title | X-Stream: Exploring MLLMs as Multiplexers for Multi-Stream Understanding |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2606.02482 |