VGenST-Bench: A Benchmark for Spatio-Temporal Reasoning via Active Video Synthesis
For researchers evaluating MLLMs, this benchmark addresses the limitation of passive video curation by providing actively synthesized, diverse scenarios for precise spatio-temporal reasoning assessment.
VGenST-Bench introduces a benchmark for spatio-temporal reasoning in MLLMs using generative models to actively synthesize controlled video scenarios, enabling fine-grained evaluation. The benchmark includes a 3x2x2 taxonomy and hierarchical tasks, with human quality control ensuring data quality.
Spatio-temporal reasoning is a core capability for Multimodal Large Language Models (MLLMs) operating in the real world. As such, evaluating it precisely has become an essential challenge. However, existing spatio-temporal reasoning benchmark datasets primarily rely on static image sets or passively curated video data, which limits the evaluation of fine-grained reasoning capabilities. In this paper, we introduce VGenST-Bench, a video benchmark that employs generative models to actively synthesize highly controlled and diverse evaluation scenarios. To construct VGenST-Bench, we propose a multi-agent pipeline incorporating a human quality control stage, ensuring the quality of all generated videos and QA pairs. We establish a comprehensive 3x2x2 video taxonomy, encompassing Spatial Scale, Perspective, and Scene Dynamics to span diverse scenarios. Furthermore, we design a hierarchical task suite that decouples low-level visual perception from high-level spatio-temporal reasoning. By shifting the paradigm from passive curation to active synthesis, VGenST-Bench enables fine-grained diagnosis of spatio-temporal understanding in MLLMs.