RePrompt: Reasoning-Augmented Reprompting for Text-to-Image Generation via Reinforcement Learning

TL;DR

RePrompt introduces a reinforcement learning-based framework that enhances text prompts for image generation by explicitly reasoning about visual semantics, leading to more accurate and faithful image synthesis.

Contribution

It is the first to incorporate explicit reasoning into prompt enhancement for text-to-image models using reinforcement learning, improving fidelity and compositionality.

Findings

Significantly improves spatial layout fidelity.

Enhances compositional generalization across models.

Achieves new state-of-the-art results on benchmarks.

Abstract

Despite recent progress in text-to-image (T2I) generation, existing models often struggle to faithfully capture user intentions from short and under-specified prompts. While prior work has attempted to enhance prompts using large language models (LLMs), these methods frequently generate stylistic or unrealistic content due to insufficient grounding in visual semantics and real-world composition. Inspired by recent advances in reasoning for language model, we propose RePrompt, a novel reprompting framework that introduces explicit reasoning into the prompt enhancement process via reinforcement learning. Instead of relying on handcrafted rules or stylistic rewrites, our method trains a language model to generate structured, self-reflective prompts by optimizing for image-level outcomes. The tailored reward models assesse the generated images in terms of human preference, semantic alignment, and visual composition, providing indirect supervision to refine prompt generation. Our approach enables end-to-end training without human-annotated data. Experiments on GenEval and T2I-Compbench show that RePrompt significantly boosts spatial layout fidelity and compositional generalization across diverse T2I backbones, establishing new state-of-the-art results.