L3GO: Language Agents with Chain-of-3D-Thoughts for Generating Unconventional Objects
This addresses a specific limitation in AI for generating out-of-distribution 3D objects, offering a novel inference-time method that is incremental in improving reasoning capabilities.
The paper tackles the problem of generating unconventional objects with precise physical and spatial configurations, which current diffusion models struggle with, by proposing L3GO, a language agent that uses chain-of-3D-thoughts for 3D mesh generation, achieving superior performance over GPT-4 and other agents on ShapeNet and a new UFO benchmark.
Diffusion-based image generation models such as DALL-E 3 and Stable Diffusion-XL demonstrate remarkable capabilities in generating images with realistic and unique compositions. Yet, these models are not robust in precisely reasoning about physical and spatial configurations of objects, especially when instructed with unconventional, thereby out-of-distribution descriptions, such as "a chair with five legs". In this paper, we propose a language agent with chain-of-3D-thoughts (L3GO), an inference-time approach that can reason about part-based 3D mesh generation of unconventional objects that current data-driven diffusion models struggle with. More concretely, we use large language models as agents to compose a desired object via trial-and-error within the 3D simulation environment. To facilitate our investigation, we develop a new benchmark, Unconventionally Feasible Objects (UFO), as well as SimpleBlenv, a wrapper environment built on top of Blender where language agents can build and compose atomic building blocks via API calls. Human and automatic GPT-4V evaluations show that our approach surpasses the standard GPT-4 and other language agents (e.g., ReAct and Reflexion) for 3D mesh generation on ShapeNet. Moreover, when tested on our UFO benchmark, our approach outperforms other state-of-the-art text-to-2D image and text-to-3D models based on human evaluation.