用事件触发机制实现长视频实时理解，避免重复与信息丢失。
Event-VStream: Event-Driven Real-Time Understanding for Long Video Streams
- 以语义事件为单位处理视频，仅在关键变化处生成语言
- 在2小时视频上保持70% GPT-5胜率，性能接近Flash-VStream
- 适合需要低延迟长时推理的实时视频分析场景
长视频流的实时理解对多模态大语言模型仍具挑战，源于冗余帧处理和历史上下文快速遗忘。现有系统依赖固定间隔解码或缓存剪枝，导致输出重复或丢失关键时间信息。我们提出Event-VStream，一种事件感知框架，将连续视频表示为一系列语义连贯的离散事件。通过融合运动、语义与预测线索检测有意义的状态转换，并仅在边界触发语言生成。每个事件嵌入被整合至持久记忆库，支持长时推理且保持低延迟。在OVOBench-Realtime与长视频Ego4D评估中，Event-VStream表现优异：相比VideoLLM-Online-8B基准提升10.4分，在2小时Ego4D流上维持约70% GPT-5胜率，性能接近Flash-VStream-7B，仅使用通用的LLaMA-3-8B文本骨干。
原文摘要 · Abstract (English)
Real-time understanding of long video streams remains challenging for multimodal large language models (VLMs) due to redundant frame processing and rapid forgetting of past context. Existing streaming systems rely on fixed-interval decoding or cache pruning, which either produce repetitive outputs or discard crucial temporal information. We introduce Event-VStream, an event-aware framework that represents continuous video as a sequence of discrete, semantically coherent events. Our system detects meaningful state transitions by integrating motion, semantic, and predictive cues, and triggers language generation only at those boundaries. Each event embedding is consolidated into a persistent memory bank, enabling long-horizon reasoning while maintaining low latency. Across OVOBench-Realtime, and long-form Ego4D evaluations, Event-VStream achieves competitive performance. It improves over a VideoLLM-Online-8B baseline by +10.4 points on OVOBench-Realtime, achieves performance close to Flash-VStream-7B despite using only a general-purpose LLaMA-3-8B text backbone, and maintains around 70% GPT-5 win rate on 2-hour Ego4D streams.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。