Scene-Designer: Learning Design Intelligence for Agentic 3D Scene Generation

Key points

• Generating complete, coherent, and usable 3D indoor scenes from natural language requires strong design intelligence: the ability to design high-quality scenes and critique their quality.1
• We construct two datasets: Scene-OTA-35K, which provides Observation--Thought--Action supervision for learning scene design and efficient error correction, and Scene-OTC-15K, which provides Observation--Thought--Critic supervision for learning scene quality critique across physical validity, functional correctness, and layout coherence.1
• Based on these experiences, we propose Scene-GRPO, a Critic-guided GRPO framework that improves design decisions and reduces unnecessary trial-and-error corrections.1
• Extensive experiments demonstrate that Scene-Designer achieves superior performance in physical validity, functional correctness, and layout coherence, while the learned Scene-Critic significantly outperforms general-purpose models in scene critique.1

Sources (2)

• [1]Scene-Designer: Learning Design Intelligence for Agentic 3D Scene Generation
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 16, 08:28 AM

  In this work, we introduce Scene-Designer, an agentic framework that learns design intelligence through complementary design and critique experiences.

  Generating complete, coherent, and usable 3D indoor scenes from natural language requires strong design intelligence: the ability to design high-quality scenes and critique their quality.

• [2]Scene-Designer: Learning Design Intelligence for Agentic 3D Scene Generation
  Hugging Face Daily Papers · Oct 16, 12:00 AM · same content

Extractive summary: sentences quoted from the sources.