{"slug": "event-stvg-spatio-temporal-video-grounding-for-long-form-event-reasoning", "title": "Event-STVG: Spatio-Temporal Video Grounding for Long-Form Event Reasoning", "summary": "Researchers at Tsinghua University's THU-KEG introduced Event-STVG, a spatio-temporal video grounding benchmark of 4,016 annotated event occurrences across soccer and seven additional domains, shifting the task from instance-specific localization to long-form event reasoning. In experiments with state-of-the-art multimodal large language models, the strongest model reached only 5.4% m-vIoU at the 10-minute horizon, with the authors reporting consistent confusion among repeated same-type events, spatial grounding bottlenecks, and rare success in jointly grounding all required roles. Code and annotations are available at https://github.com/THU-KEG/EventSTVG.", "body_md": "arXiv:2610.22444v1 Announce Type: new \nAbstract: Spatio-Temporal Video Grounding (STVG) aims to jointly localize a target in space and time given a natural-language query. Despite continued advances, existing STVG benchmarks remain largely centered on localizing targets through instance-specific descriptions, leaving them largely decoupled from long-form reasoning increasingly emphasized in multimodal large language models (MLLMs). In this paper, we introduce Event-STVG, a task that shifts STVG from locating what is described toward event-level reasoning by unifying long-form context, repeated-event reasoning, event localization, role inference, and multi-target grounding. To facilitate research on Event-STVG, we construct a large-scale benchmark comprising 4,016 annotated event occurrences across soccer and seven additional domains. Experiments with state-of-the-art MLLMs reveal substantial difficulty, with the strongest model achieving only 5.4% m-vIoU at the 10-minute horizon. Across domains, we consistently observe confusion among repeated same-type events, spatial grounding bottlenecks, and rare success in jointly grounding all required roles. Event-STVG establishes a new challenge for visual understanding in MLLMs, calling for future models that can jointly reason about complex events across space and time. Code and annotations are available at https://github.com/THU-KEG/EventSTVG", "url": "https://wpnews.pro/news/event-stvg-spatio-temporal-video-grounding-for-long-form-event-reasoning", "canonical_source": "https://arxiv.org/abs/2610.22444", "published_at": "2026-10-19 04:00:00+00:00", "updated_at": "2026-10-19 04:21:41.483633+00:00", "lang": "en", "topics": ["computer-vision", "ai-research", "large-language-models", "machine-learning"], "entities": ["Event-STVG", "THU-KEG", "Tsinghua University", "GitHub"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/event-stvg-spatio-temporal-video-grounding-for-long-form-event-reasoning", "markdown": "https://wpnews.pro/news/event-stvg-spatio-temporal-video-grounding-for-long-form-event-reasoning.md", "text": "https://wpnews.pro/news/event-stvg-spatio-temporal-video-grounding-for-long-form-event-reasoning.txt", "jsonld": "https://wpnews.pro/news/event-stvg-spatio-temporal-video-grounding-for-long-form-event-reasoning.jsonld"}}