Search papers, labs, and topics across Lattice.
This paper introduces SVI-Bench, a comprehensive benchmark designed to evaluate Strategic Video Intelligence (SVI) by simulating complex multi-agent interactions in team sports. By combining real-world dynamics with verifiable outcomes, SVI-Bench provides a unique dataset comprising 35K hours of video and extensive annotations, enabling a structured assessment of capabilities from dynamic scene understanding to strategic planning. The findings reveal a significant performance drop in models as tasks progress from basic perception to more complex agentic synthesis, highlighting a critical gap in current AI capabilities for strategic reasoning in dynamic environments.
Models excel at recognizing actions in sports videos but falter dramatically when tasked with strategic reasoning, achieving only 5% accuracy in autonomous evidence integration.
True video intelligence demands more than recognizing what is visible: it requires reasoning about why events unfold, predicting what would change under different conditions, and deciding what to do next. We refer to this progression, from perception through causal reasoning and simulation to strategic planning, as Strategic Video Intelligence (SVI). No existing benchmark evaluates this capability stack: in-the-wild videos lack verifiable ground truth for causal and strategic questions, while synthetic environments sacrifice the complexity of real multi-agent systems. To bridge this gap, we introduce SVI-Bench, a large-scale benchmark that leverages team sports as a dynamic microworld, combining the complexity of real-world multi-agent interaction (10-22 agents making coordinated decisions under adversarial pressure) with the verifiability of explicit rules and definitive outcomes. SVI-Bench comprises approximately 35K hours of broadcast video, 15M annotated actions, 15K hours of expert commentary, 23K game reports, and 103K structured statistical records across basketball, soccer, and hockey, all constructed via a data engine that transforms raw game data into a dense, cross-referenced corpus. We organize evaluation into 9 tasks spanning a progressive four-pillar hierarchy: Dynamic Scene Understanding, Causal Reasoning, Strategic Simulation, and Agentic Synthesis. Evaluating strong multimodal and agentic baselines, we find a capability cliff: models perform competently on perceptual tasks, achieving approximately 73% on fine-grained action QA, but degrade sharply at each successive cognitive level. Agentic tasks prove hardest: the strongest model achieves only 5% accuracy when required to autonomously gather and integrate evidence across a corpus of 1.8M clips.
