aiqbench analysis
This research introduces LEGO, a new benchmark and evaluation framework designed to rigorously test text-guided 3D scene synthesis by verifying fine-grained spatial constraints rather than relying on coarse similarity metrics. The study reveals a significant performance gap, showing that current generative approaches succeed in only about 10% of these complex verification tasks. This work is critical for advancing embodied AI agents that must accurately interpret and render detailed textual instructions into coherent 3D environments.
Apples on the Table? Evaluating Text-Guided 3D Scene Synthesis via Fine-Grained Constraint Verification
arXiv:2511.03001v3 Announce Type: replace Abstract: Accurately synthesizing 3D scenes from user-provided text descriptions is crucial for developing embodied agents. Despite the importance of scene-description alignment, existing evaluation methods for such text-guided 3D scene synthesis either capture only coarse similarity between the synthesized scene and the user description, or ignore the spatial reasoning for verifying object placement. None of them addressed the fine-grained constraints (e.g., X needs to be in the scene in a Y manner) implied by the description from users. To address this, we introduce LEGO, a benchmark dataset that pairs each user description with human-annotated constraints and a reference scene, and LEGO-Eval, an evaluation framework that decomposes a description into atomic constraints and verifies each one using tools that ground textual references to 3D objects and reason about their spatial relationships. We show that (i) LEGO-Eval evaluates misalignment far more accurately than existing methods and (ii) current scene synthesis approaches achieve at most 10% success rate in LEGO-Eval.
Title:Apples on the Table? Evaluating Text-Guided 3D Scene Synthesis via Fine-Grained Constraint Verification
Abstract:Accurately synthesizing 3D scenes from user-provided text descriptions is crucial for developing embodied agents. Despite the importance of scene-description alignment, existing evaluation methods for such text-guided 3D scene synthesis either capture only coarse similarity between the synthesized scene and the user description, or ignore the spatial reasoning for verifying object placement. None of them addressed the fine-grained constraints (e.g., X needs to be in the scene in a Y manner) implied by the description from users. To address this, we introduce LEGO, a benchmark dataset that pairs each user description with human-annotated constraints and a reference scene, and LEGO-Eval, an evaluation framework that decomposes a description into atomic constraints and verifies each one using tools that ground textual references to 3D objects and reason about their spatial relationships. We show that (i) LEGO-Eval evaluates misalignment far more accurately than existing methods and (ii) current scene synthesis approaches achieve at most 10% success rate in LEGO-Eval.
Computation and Language (cs.CL)
arXiv:2511.03001 [cs.CL]
(or arXiv:2511.03001v3 [cs.CL] for this version)
https://doi.org/10.48550/arXiv.2511.03001 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 261 words
