SceneRAG uses LLM-driven scene segmentation and a scene-level knowledge graph to retrieve and answer questions about long videos, reporting higher LLM-judged win-rates than chunk-based RAG baselines on the LongerVideos benchmark.
Unified graph structured models for video understanding
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
SceneRAG: Scene-level Retrieval-Augmented Generation for Video Understanding
SceneRAG uses LLM-driven scene segmentation and a scene-level knowledge graph to retrieve and answer questions about long videos, reporting higher LLM-judged win-rates than chunk-based RAG baselines on the LongerVideos benchmark.