OASIS：面向流式视频推理的按需分层事件记忆
OASIS: On-Demand Hierarchical Event Memory for Streaming Video Reasoning
把无限增长的流式视频历史组织成分层事件结构，并用「先短上下文推理、不确定时再按高层意图检索」的按需检索策略替代稠密帧采样或嵌入相似度检索，可直接迁移到长视频问答与流式 MLLM 的记忆管理模块。
原文依据 performs reasoning as controlled refinement -- short-context inference first, followed by semantically grounded retrieval only when uncertainty arises
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论流式视频推理要求模型在历史无界增长、而有意义的证据始终稀缺的设定下运行。在这样的图景中，相关信号如同沙漠中的绿洲——微小、关键，且极易淹没在冗余之中。单纯扩大记忆只会让沙漠更宽，而激进压缩则会抽干绿洲。真正的困难在于发现该往哪里看，而非记住多少。为此，我们提出 OASIS，一个面向流式视频推理的新框架，通过结构化的按需检索来应对这一挑战。它将流式历史组织为分层事件，并把推理实现为受控的细化过程——先进行短上下文推理，仅在出现不确定性时才执行语义接地的检索。由于检索由高层意图而非嵌入相似度驱动，所检索到的记忆显著更准确、噪声更少。此外，该机制即插即用、无需训练，可便捷地挂接到不同的流式多模态大模型骨干上。在多个基准与骨干上的实验表明，OASIS 在有界词元开销与低请求延迟下，于长时程准确率和组合式推理上取得显著提升。
英文原文摘要 1244 字符
Streaming video reasoning requires models to operate in a setting where history grows without bound while meaningful evidence remains scarce. In such a landscape, relevant signal is like an oasis -- small, critical, and easily lost in a desert of redundancy. Enlarging memory only widens the desert; aggressive compression dries up the oasis. The real difficulty lies in discovering where to look, not how much to remember. We therefore introduce OASIS, a novel framework for streaming video reasoning that tackles this challenge through structured, on-demand retrieval. It organizes streaming history into hierarchical events and performs reasoning as controlled refinement -- short-context inference first, followed by semantically grounded retrieval only when uncertainty arises. As the retrieval is driven by high-level intent rather than embedding similarity, the retrieved memory is substantially more accurate and less noisy. Additionally, the mechanism is plug-and-play, training-free, and readily attaches to different streaming MLLM backbones. Experiments across multiple benchmarks and backbones show that OASIS achieves strong gains in long-horizon accuracy and compositional reasoning with bounded token cost and low request delay.