SemanticKV：面向智能体自回归视频生成的语义对齐 KV 缓存压缩
Semantically Aligned KV Cache Compression for Efficient Agentic Autoregressive Video Generation
做长视频生成或智能体视频合成的开发者可以看看这篇。它不用重训练，靠规划器的语义约束决定缓存留什么，60 秒生成时内存砍到三分之一，画质反而比不压缩还高一点。四个骨干、三个基准都有数据，方法可直接套在现有 rollout 流程里。
论文提出 SemanticKV，一个免训练的 KV 缓存压缩方法，专门针对结合语义规划、顺序生成与视觉验证的智能体自回归视频生成场景。据作者所述，这是首个显式利用规划器提取的语义约束来引导缓存保留策略的压缩方法，解决现有方法只优化单个生成器、可能导致支持目标属性或动作的状态被低优先级丢弃的问题。
技术上，方法通过文本交叉注意力定位与语义约束相关的视频位置，将这些位置的自注意力查询分组，组内平均注意力权重、跨组取最大值打分，再用一次 top-k 选择在内存预算内保留高分条目，键、值与位置保持不变。作者证明在查询组固定、条目成本相等的条件下，该选择能最小化最差组被丢弃注意力的加性上界。方法在每个候选 rollout 内独立运行，无需重训练或额外视频生成。
实验覆盖四个 AR 视频骨干和三个基准，测试 5、30、60 秒生成。在 Causal Forcing 上压缩比 0.7 时，SemanticKV 的 VBench Total Score 为 84.10，高于最强压缩基线的 84.05 和 Full KV 的 83.99；相对 Full KV 取得 1.22 倍加速，KV 缓存内存从 6.04 GB 降至 1.81 GB，减少 3.35 倍。
Semantically Aligned KV Cache Compression for Efficient Agentic Autoregressive Video Generation
Agentic autoregressive video generation combines semantic planning, sequential generation, and visual verification for controllable synthesis. Growing visual histories enlarge the key-value (KV) cache, increasing memory use and attention costs across candidate rollouts. Existing compression methods primarily optimize individual generators without the planner's semantic constraints, risking low priority for states supporting requested attributes or actions. We introduce SemanticKV, a training-free method for semantically aligned KV cache compression. To our knowledge, SemanticKV is the first KV cache compression method for agentic autoregressive video generation that explicitly uses planner-extracted semantic constraints to guide retention. Text cross-attention identifies constraint-related video positions, whose self-attention queries are collected into separate groups. Attention weights are averaged within each group and maximized across groups, giving each cached entry a high score when strongly used by any represented constraint. A single top-k selection retains high-scoring entries under the memory budget, leaving their keys, values, and positions unchanged. For fixed query groups and equal entry costs, this selection minimizes an additive upper bound on worst-group discarded attention. The method operates independently within each candidate without retraining or additional video generations. We evaluate four AR video backbones on three benchmarks for 5-, 30-, and 60-second generation. On Causal Forcing at a compression ratio of 0.7, SemanticKV obtains a VBench Total Score of 84.10, compared with 84.05 for the strongest compressed baseline and 83.99 for Full KV. It achieves a 1.22x speedup over Full KV and reduces KV-cache memory from 6.04 GB to 1.81 GB, a 3.35x reduction.
本条情报归属 AITOP「论文」核心追踪栏目。在跨区域产业观察坐标中，该动态直接关联「AI 开发者工具链、Coding Agent 与工程化实践」研究主线，反映出当前技术演进与商业落地的最新风向。涉及核心实体与技术议题包括：SemanticKV、视频生成、KV缓存、推理优化，读者可通过下方标签进入对应专题页查阅历史脉络与关联对比。
信源核验说明：本文原始线索采编自「arXiv cs.LG」（@Weizhong Huang, Huaijia Kou, Tianyu Xie, Xiawu Zheng, Liujuan Cao），经 AITOP 多维评分管线结构化提炼。