Generation Navigator: A State-Aware Agentic Framework for Image Generation

TL;DR

This paper introduces Generation Navigator, a state-aware agentic framework for image generation that learns to adaptively steer the process through multi-turn interactions, improving fidelity to user intent.

Contribution

It proposes a novel reinforcement learning approach, PRE-GRPO, to train a dynamic, multi-turn T2I agent that effectively balances image quality, trajectory retention, and turn efficiency.

Findings

Achieves a WISE score of 0.90 on benchmarks.

Reaches 79.06% reasoning accuracy on T2I-ReasonBench.

Substantially outperforms existing methods in multi-turn image generation.

Abstract

Despite rapid advances in text-to-image generation, faithfully realizing user intent remains challenging, often requiring manual multi-turn trial and error. To automate this process, existing systems rely on either simple prompt rewriting or closed-loop agents driven by hand-crafted rules, rather than learning to adapt actions to the evolving generation process. In this paper, we reformulate image generation as a state-conditioned action-making problem and propose Generation Navigator, a multi-turn T2I agent that learns to dynamically steer the generation trajectory and output the next action. However, training this agent via reinforcement learning introduces a critical credit assignment challenge: naively rewarding a trajectory based solely on a single state assigns equal credit to all actions in the rollout, ignores the quality dynamics across turns, and fails to distinguish actions that improve the trajectory from those that degrade it or waste turns without progress. We resolve this with PRE-GRPO (Peak-Retention-Efficiency Group Relative Policy Optimization), a trajectory-level reinforcement learning objective that explicitly rewards discovering a high-quality image (Peak), avoiding subsequent quality degradation across turns (Retention), and minimizing unnecessary turns (Efficiency). Experiments show substantial improvements across benchmarks, reaching a WISE score of 0.90 and 79.06% reasoning accuracy on T2I-ReasonBench.