论文提出 ObjectStream，用目标当锚点压缩视频记忆，免训练提升理解，显存延迟减半，性能还涨 10 分。
ObjectStream 是面向流式视频理解的无训练框架，把潜在目标作为记忆锚点。它直接从冻结的 Video-LLM 表征中提取空间连贯的潜在目标，跨帧链接为持久锚点，无需外部检测器或分割模型。在 OVO-Bench Real-Time Visual Perception 上，ObjectStream 使 Qwen2.5-VL-7B 提升 10.0 分，峰值 GPU 内存和 TTFT 均降低约 50%。在离线长视频基准上，它超越全 token 基线，同时丢弃 82.5% 的视觉 token。
ObjectStream: Latent Objects as Memory Anchors for Streaming Video Understanding
Streaming video understanding requires models to continuously retain useful visual evidence before future questions are known. Existing approaches primarily manage the growing visual context according to token importance, temporal redundancy, or segment-level relevance, but rarely organize evidence around objects that persist and evolve over time. Thus, in this paper, we introduce ObjectStream, a training-free framework that treats latent objects as memory anchors for streaming video understanding. ObjectStream induces spatially coherent latent objects directly from frozen Video-LLM representations, links them across frames into persistent anchors, and maintains their histories under a bounded memory budget, without requiring external object detectors or segmentation models. Built on these anchors, ObjectStream preserves three complementary forms of evidence: persistent object histories, transient object changes, and recent visual context. This design enables existing Video Large Language Models (Video-LLMs) to reason over object identities, interactions, and state changes while leaving the underlying model unchanged. Extensive experiments on online streaming and offline long-video benchmarks demonstrate both effectiveness and efficiency. In online streaming evaluation, ObjectStream improves Qwen2.5-VL-7B by 10.0 points on OVO-Bench Real-Time Visual Perception, while reducing peak GPU mem-ory and TTFT by approximately 50%. On offline long-video benchmarks, it surpasses the full-token baseline while discarding 82.5% of visual tokens. These results highlight latent objects as a practical and effective organizing principle for compact streaming video memory.