新流程使用 Qwen3.5-27B 进行面向推理的视频检索

作者 PulseAugur 编辑部 · [2 个来源] · 2026-06-01 16:21

研究人员开发了一种新颖的零样本面向推理的组合视频检索（CoVR-R）流程。该系统利用 Qwen3.5-27B 模型，根据应用于参考视频的编辑指令来推断目标视频。该方法结合了密集检索和 TF-IDF 分支，在验证集和测试集上取得了强劲的性能，R@1 等指标最高达到 89.73%。 AI

影响引入了一种利用推理能力进行视频检索的新颖方法，有望提高复杂视频编辑的搜索准确性。

排序理由这是一篇详细介绍视频检索新方法和模型的学术论文。

AI 生成摘要 · Google Gemini · 来自 2 个来源。我们如何撰写摘要 →

报道来源 [2]

arXiv cs.CV TIER_1 English(EN) · DongQing Liu, MengShi Qi, HongWei Ji · 2026-06-02 04:00

Reason-Then-Retrieve for CoVR-R with Structured Edit Prompts and Dense-Sparse Fusion

arXiv:2606.02450v1 Announce Type: new Abstract: CoVR-R studies reason-aware composed video retrieval: given a reference video and an edit instruction, the system must retrieve the target video that satisfies the edit. The main difficulty is that the target is not described direct…
arXiv cs.CV TIER_1 English(EN) · HongWei Ji · 2026-06-01 16:21

Reason-Then-Retrieve for CoVR-R with Structured Edit Prompts and Dense-Sparse Fusion

CoVR-R studies reason-aware composed video retrieval: given a reference video and an edit instruction, the system must retrieve the target video that satisfies the edit. The main difficulty is that the target is not described directly; it must be inferred from fine-grained change…