By measuring text ambiguity, mapping uncertainty, and frame quality, UMIVR asks targeted clarifying questions and raises Recall@1 on MSR-VTT-1k to 69.2% after 10 interaction rounds.
Chen and William B
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Quantifying and Narrowing the Unknown: Interactive Text-to-Video Retrieval via Uncertainty Minimization
By measuring text ambiguity, mapping uncertainty, and frame quality, UMIVR asks targeted clarifying questions and raises Recall@1 on MSR-VTT-1k to 69.2% after 10 interaction rounds.