#machine learning
Jun 2026
Driving Video Retrieval for Complex Queries with Structured Grounding
STRIVE-D, a data-calibrated retrieval framework for driving videos, uses weakly labeled in-domain videos to estimate when a query rule is reliable, adapt rules that mismatch observed data, and fuse calibrated rule scores with vision-language and keyword-based retrieval signals.
Man-Yi Yao, Sparsh Garg, Christian R. Shelton et al.
· arXiv.org · 1 citation