RS-Gen: A Multi-Stage Agentic Framework for Reasoning and Search-Augmented Image Generation

RS-Gen is a novel multi-stage agentic framework designed for reasoning and search-augmented image generation, addressing limitations in existing models when faced with ambiguous intentions and Out-of-Distribution knowledge. It features a "Questioning-and-Solving" closed-loop mechanism that enhances logical reasoning and fills knowledge gaps without the need for additional training. Experimental results show RS-Gen achieves significant performance improvements on the WISE Verified and RISEBench benchmarks, elevating the Qwen-Image and Qwen-Image-Edit-2511 models to state-of-the-art status among open-source solutions, which is crucial for practitioners seeking to enhance image generation capabilities in dynamic contexts.