VAM is a training-free framework with online indexing, hierarchical memory in parallel representations, and agentic retrieval that improves long video understanding on OVO-Bench and month-scale MM-Lifelong splits over baseline MLLM use.
Tool bo un dar y rules : 19 - ’ search ’ is the default tool for a n s w e r i n g q u e s t i o n s
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Visual Agentic Memory: Enabling Online Long Video Understanding via Online Indexing, Hierarchical Memory, and Agentic Retrieval
VAM is a training-free framework with online indexing, hierarchical memory in parallel representations, and agentic retrieval that improves long video understanding on OVO-Bench and month-scale MM-Lifelong splits over baseline MLLM use.