Search papers, labs, and topics across Lattice.
The paper introduces OVO-S-Bench, a novel benchmark designed to evaluate streaming spatial intelligence in multimodal large language models (LLMs) through a comprehensive set of 1,680 human-annotated questions derived from 348 source videos. This benchmark assesses models on four levels of spatial reasoning, revealing that even advanced models like Gemini-3.1-Pro significantly lag behind human performance, particularly in allocentric mapping, which is identified as a critical bottleneck. The findings underscore the limitations of current streaming and spatially fine-tuned LLMs, highlighting the challenges of chain-of-thought reasoning when ungrounded in real-time data.
Streaming spatial intelligence remains a significant hurdle for multimodal LLMs, with top models trailing human experts by 27 points in allocentric mapping tasks.
Multimodal agents in robotics, AR, and autonomous driving must reason about places and layouts from continuous egocentric streams, often using evidence outside the current view. Existing benchmarks either evaluate offline over full videos or target events rather than spatial structure. We introduce OVO-S-Bench, a fully human-annotated benchmark for streaming spatial intelligence, comprising 1,680 questions over 348 source videos. Annotation involves 12 trained annotators, each also serving as a blind cross-reviewer, across roughly 804 person-hours of multi-round quality assurance. Each question carries a query timestamp and an evidence interval, and at evaluation, the model sees only the prefix preceding the query. Questions span four levels of increasing abstraction: instantaneous egocentric perception, spatiotemporal context tracking, spatial simulation and reasoning, and allocentric mapping. Across 38 proprietary and open-source MLLMs, Gemini-3.1-Pro trails human experts by 27 points, 59.2 vs. 86.6, with allocentric mapping as the dominant bottleneck. Notably, streaming and spatially fine-tuned MLLMs underperform their own backbones. We further find that chain-of-thought reasoning amplifies spatial errors when ungrounded in the stream. By exposing these limitations, OVO-S-Bench establishes a demanding testbed for next-generation streaming spatial MLLMs.
Tsinghua AI