Search papers, labs, and topics across Lattice.
To resolve the trade-off between embedding-based retrieval throughput and heavy multimodal reasoning in composed video retrieval (CoVR), this work establishes CoVRAGE, a training-free framework that dynamically orchestrates frozen foundation models according to query ambiguity. The pipeline adapts inference depth through a multi-stage hierarchy: initial screening via reusable video embeddings, LLM-driven target-description generation for ambiguous edits, and stage-specific frame-rate and resolution tuning during final multimodal candidate verification. On full-gallery benchmarks, CoVRAGE establishes a new training-free state of the art, achieving 89.55 R@1 on Dense-WebVid-CoVR and 93.43 R@1 on CoVR-R, surpassing prior training-free baselines by massive 35% and 25% absolute margins.
Zero-shot composed video retrieval can jump by over 35% absolute R@1 without any parameter updates鈥攑rovided you route frozen foundation models as a dynamic compute ladder rather than a uniform scoring function.
Composed video retrieval (CoVR) searches a gallery for the target video that realizes a natural-language modification of a source clip. However, at gallery scale, this creates a fundamental tension: compact embeddings enable efficient, reusable search but can miss the transient actions, state changes, and subtle constraints that demand fine-grained video reasoning, whereas applying large multimodal models uniformly sacrifices scalability. To address these limitations, we propose that frozen foundation models should instead occupy complementary roles, with inference depth adapted to query difficulty. Based on this premise, we introduce \methodname{}, a framework for training-free \methodexpansion{}. Specifically, a composed-query embedding first searches reusable video-only gallery representations; uncertain queries undergo bounded reranking and candidate expansion; ambiguous edits trigger target-description generation; and only close leading candidates reach multimodal verification. To support these roles, frame selection, spatial resolution, and time cues are adapted to each stage. Across complete target-gallery evaluations, our method reaches state-of-the-art performance among training-free approaches, with 89.55 and 93.43 R@1 on Dense-WebVid-CoVR and CoVR-R, respectively (with more than +35\% and +25\% absolute margins to the closest counterpart). These results show that adaptively orchestrating foundation-model capabilities can combine scalable retrieval with fine-grained reasoning without task-specific training. The source code and all relevant guidelines are available on https://github.com/demidovd98/CoVRAGE.
