iARCS: Iterative Agentic RL for Controllable 3D Scene Generation. DDPO fine-tuning of a 3D indoor scene diffusion model with Eureka-style LLM rewards. arXiv 2608.06161
reinforcement-learning computer-vision rl eureka lora 3d synthetic-data diffusion-models embodied-ai 3d-scene-generation rlhf agentic-workflow indoor-scene-synthesis agentic-ai ddpo 3d-front scene-layout-generation llm-reward-design
-
Updated
Oct 6, 2026 - Python