记忆很重要：用可学习查找表提升免训练零样本时序动作定位
Memory Matters: Boosting Training-Free Zero-Shot Temporal Action Localization with a Learnable Lookup Table
把跨视频的历史知识以可学习查找表的形式累积起来，替代逐视频独立的测试时自适应，可直接迁移到其他免训练 VLM 视频/图像任务（如零样本动作识别、开放词表检测）中做在线记忆增强。
原文依据 current techniques perform test-time adaptation independently on each video, neglecting the potential benefit of accumulating knowledge from historical test videos
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论零样本时序动作定位（ZS-TAL）旨在对训练阶段未见过的动作，在未裁剪视频中进行分类与定位。现有的基于训练的 ZS-TAL 方法通常依赖在大规模标注训练数据上微调模型。这在实际应用中往往不可行，并且会损害模型的泛化能力。因此，免训练 ZS-TAL 受到关注，它直接利用视觉语言模型（VLM），无需任何额外训练即可实现动作定位。然而，当前技术对每个视频独立地进行测试时自适应，忽视了从历史测试视频中累积知识的潜在收益。为解决这一问题，我们提出了一种可学习查找表（LLT）框架。在测试过程中，我们通过纳入高置信度、多样化的查找候选来持续更新查找表，以构建动作正例查找项。此外，我们引入一个可学习残差模块，使相应的查找项适应当前视频的上下文特征。最后，我们利用精炼后的激活分数来选择准确的视频帧，并进一步调整文本原型。这种简单而有效的文本-视觉协作使免训练 ZS-TAL 能够利用历史视频。大量实验表明，我们的方法显著优于最优的零样本 VLM 基线，验证了该框架的有效性。
英文原文摘要 1420 字符
Zero-Shot Temporal Action Localization (ZS-TAL) aims to classify and localize actions in untrimmed videos that are unseen during training. Existing training-based ZS-TAL methods typically rely on fine-tuning models on large-scale annotated training data. This can be impractical in real-world applications and damage its generalization. As a result, Training-Free ZS-TAL has gained attention, which directly leveraging Vision-Language Models (VLM) enables action localization without any additional training. However, current techniques perform test-time adaptation independently on each video, neglecting the potential benefit of accumulating knowledge from historical test videos. To address this, we propose a learnable lookup table (LLT) framework. During testing, we continuously update lookup table by incorporating high-confidence, diverse lookup candidates to construct action-positive lookup item. Additionally, we introduce a learnable residual module to adapt the corresponding lookup item to the current video context features. Finally, we employ refined activation scores to select accurate video frames and further adjust the text prototypes. This simple yet effective text-visual collaboration enables training-free ZS-TAL to harness historical videos. Extensive experiments show our method significantly outperforms state-of-the-art zero-shot VLM baselines, validating the effectiveness of our framework.