CRAFT: Continuous Reasoning and Agentic Feedback Tuning for Multimodal Text-to-Image Generation
This work addresses reliability issues in multimodal generative models for users needing interpretable and controllable image generation, representing an incremental advance by adapting structured reasoning from language models to this domain.
The paper tackles the problem of improving text-to-image generation by introducing CRAFT, a training-free framework that uses structured reasoning to decompose prompts, verify images, and apply targeted edits, resulting in consistent improvements in compositional accuracy and text rendering across multiple models with negligible overhead.
Recent work has shown that inference-time reasoning and reflection can improve text-to-image generation without retraining. However, existing approaches often rely on implicit, holistic critiques or unconstrained prompt rewrites, making their behavior difficult to interpret, control, or stop reliably. In contrast, large language models have benefited from explicit, structured forms of **thinking** based on verification, targeted correction, and early stopping. We introduce CRAFT (Continuous Reasoning and Agentic Feedback Tuning), a training-free, model-agnostic framework that brings this structured reasoning paradigm to multimodal image generation. CRAFT decomposes a prompt into dependency-structured visual questions, veries generated images using a vision-language model, and applies targeted prompt edits through an LLM agent only where constraints fail. The process iterates with an explicit stopping criterion once all constraints are satised, yielding an interpretable and controllable inference-time renement loop. Across multiple model families and challenging benchmarks, CRAFT consistently improves compositional accuracy, text rendering, and preference-based evaluations, with particularly strong gains for lightweight generators. Importantly, these improvements incur only a negligible inference-time overhead, allowing smaller or cheaper models to approach the quality of substantially more expensive systems. Our results suggest that explicitly structured, constraint-driven inference-time reasoning is a key ingredient for improving the reliability of multimodal generative models.