MANSION: Multi-floor lANguage-to-3D Scene generatIOn for loNg-horizon tasks

2026-03-12

AI Summary

Existing embodied AI benchmarks are largely confined to single-floor indoor environments, limiting their capacity to support spatial reasoning required for cross-floor, long-horizon tasks. To address this gap, this work proposes the first language-driven, building-scale multi-floor 3D scene generation framework that integrates architectural priors, language-to-3D generation, open-vocabulary semantic editing, and navigability constraints to enable diverse, vertically structured scene synthesis. Leveraging this framework, we introduce MansionWorld, a dataset comprising over 1,000 multi-story buildings. Benchmark evaluations demonstrate a significant performance drop in state-of-the-art agents, underscoring the dataset’s effectiveness in presenting a more challenging and evaluatively meaningful testbed for embodied intelligence.

Technology Category

• Intelligent Robots: Embodied AI
• Natural Language Processing: Language Grounding & Multi-modal NLP
• Multiagent Systems: Agent/AI Theories and Architectures

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Social Networks and Social Media: Generative AI / large language models and their impact on social systems

Abstract

Real-world robotic tasks are long-horizon and often span multiple floors, demanding rich spatial reasoning. However, existing embodied benchmarks are largely confined to single-floor in-house environments, failing to reflect the complexity of real-world tasks. We introduce MANSION, the first language-driven framework for generating building-scale, multi-floor 3D environments. Being aware of vertical structural constraints, MANSION generates realistic, navigable whole-building structures with diverse, human-friendly scenes, enabling the development and evaluation of cross-floor long-horizon tasks. Building on this framework, we release MansionWorld, a dataset of over 1,000 diverse buildings ranging from hospitals to offices, alongside a Task-Semantic Scene Editing Agent that customizes these environments using open-vocabulary commands to meet specific user needs. Benchmarking reveals that state-of-the-art agents degrade sharply in our settings, establishing MANSION as a critical testbed for the next generation of spatial reasoning and planning.

Problem

• multi-floor
• long-horizon tasks
• 3D scene generation
• spatial reasoning
• embodied AI

Innovation

• language-to-3D
• multi-floor generation
• spatial reasoning
• embodied AI
• scene editing