A new 1,000-video benchmark with manually separated spatial and temporal captions, two spatiotemporal bias metrics, and a unified MLLM baseline for captioning and retrieval.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2024 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
CaReBench: A Fine-Grained Benchmark for Video Captioning and Retrieval
A new 1,000-video benchmark with manually separated spatial and temporal captions, two spatiotemporal bias metrics, and a unified MLLM baseline for captioning and retrieval.