不训练的缓存搜索让视频生成提速三倍，还能保持高精度。
CachedSearch: Training-Free Cached Exploration for Test-Time Search in Video Diffusion
- 用无训练缓存加速测试时搜索，每轮快2-3倍
- 缓存后重算最优解，保留90%以上完整搜索收益
- 适合想低成本提升视频生成质量的研究者
测试时搜索能让小型视频扩散模型媲美大型模型，但耗时增加2-10倍。所有候选样本均完全去噪，而多数被丢弃。无训练缓存使每轮推理快2-3倍，且质量近乎无损。关键在于：缓存是否破坏验证器对候选排序的准确性。本文首次研究缓存对视频测试时搜索排名的影响。在Wan2.1-T2V-1.3B上，采用自适应缓存封装（每候选加快约2倍），ImageReward评分显示缓存与全量生成结果高度一致。中位数提示斯皮尔曼等级相关性达0.905，VBench套件上Top-1匹配率达72%。VBench-2.0在更难数据集上复现此结果。重新计算缓存最优解可保留90%-94%的全搜索增益。错误集中于近似平局的候选，因此干扰自我限制。基于此提出CachedSearch：先以激进缓存探索全部候选，再仅对胜出者进行全算力重生成。在N=8时，仅需63%成本即捕获94.7%的最佳-组合收益。探索宽度越大，捕获率越高。相同预算下，可将搜索宽度翻倍，增益提升38%。该效果在1.3B至14B的六种模型、四个架构族（Wan、LTX、CogVideoX、Hunyuan）中均成立。Wan2.1-14B在保真度上达到1.3B模型水平。中轨迹剪枝使探索效率提升至3.11倍，捕获率88.6%。跨架构移植仅需微调单一参数，表明保真度依赖架构而非参数量。CachedSearch无需训练、对验证器和搜索算法均无依赖，可作为测试时扩展的即插即用增强模块。
原文摘要 · Abstract (English)
Test-time search lets small video diffusion models rival larger ones, but costs 2-10x more. All candidates are fully denoised, although most are discarded. Training-free caching makes each rollout 2-3x faster at near-lossless quality. Composition is safe only if lossy caching preserves verifier rankings. We present the first study of whether caching corrupts candidate ranking in video test-time search. On Wan2.1-T2V-1.3B with an adaptive caching wrapper (~2x per-candidate speedup), ImageReward scores seed-matched cached and full rollouts. Median per-prompt Spearman rank correlation is 0.905, with 72% top-1 agreement on the VBench suite. VBench-2.0 replicates this result on a harder suite. Recomputing the cached winner at full compute retains 90-94% of the full-search gain. Errors cluster among near-tied candidates, making corruption self-limiting. This finding leads to CachedSearch. It explores every candidate with aggressive caching, then re-generates only the winner at full compute. At N=8, it captures 94.7% of best-of-N's gain at 63% of the cost. Capture rises with width. At matched budget, it searches twice as wide for 38% more gain. The result holds from 1.3B-14B across six models and four families: Wan, LTX, CogVideoX, and Hunyuan. Wan2.1-14B matches the 1.3B model's fidelity. Mid-trajectory pruning multiplies the exploration saving to 3.11x at 88.6% capture. Ports to other model families require recalibrating a single parameter, showing that fidelity tracks architecture rather than parameter count. CachedSearch is training-free, verifier-agnostic, and orthogonal to the search algorithm, making it a plug-in multiplier for test-time scaling.
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。