A new zero‑shot framework uses multimodal LLMs to locate spatio‑temporal tubes in video from natural‑language queries, outperforming state‑of‑the‑art methods on three benchmark datasets. https://getnews.me/multimodal-llms-enable-zero-shot-spatio-temporal-video-grounding/ #zeroshot #videogrounding

TimeLens: Rethinking Video Temporal Grounding with Multimodal LLMs
Jun Zhang, Limin Wang et al.
Paper
Details
#VideoGrounding #MultimodalAI #TemporalReasoning