T2I-R1: Reinforcing Image Generation with Collaborative Semantic-level and Token-level CoT
About
Recent advancements in large language models have demonstrated how chain-of-thought (CoT) and reinforcement learning (RL) can improve performance. However, applying such reasoning strategies to the visual generation domain remains largely unexplored. In this paper, we present T2I-R1, a novel reasoning-enhanced text-to-image generation model, powered by RL with a bi-level CoT reasoning process. Specifically, we identify two levels of CoT that can be utilized to enhance different stages of generation: (1) the semantic-level CoT for high-level planning of the prompt and (2) the token-level CoT for low-level pixel processing during patch-by-patch generation. To better coordinate these two levels of CoT, we introduce BiCoT-GRPO with an ensemble of generation rewards, which seamlessly optimizes both generation CoTs within the same training step. By applying our reasoning strategies to the baseline model, Janus-Pro, we achieve superior performance with 13% improvement on T2I-CompBench and 19% improvement on the WISE benchmark, even surpassing the state-of-the-art model FLUX.1. Code is available at: https://github.com/CaraJ7/T2I-R1
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Text-to-Image Generation | GenEval | Overall Score79 | 467 | |
| Text-to-Image Generation | T2I-CompBench | Shape Fidelity58.52 | 94 | |
| Text-to-Image Generation | GenEval | Overall Score79 | 68 | |
| Knowledge-grounded reasoning | WISE | Overall Score54 | 45 | |
| Text-to-Image Generation | DPG-Bench (test) | Global Fidelity91.79 | 43 | |
| World Knowledge Image Generation | WISE | Overall Score54 | 39 | |
| Text-to-Image Generation | GenEval++ | Color Accuracy68 | 35 | |
| Text-to-Image Generation | Imagine-Bench | Attribute Shift Score5.85 | 33 | |
| Text-to-Image Generation | WISE (test) | Overall Score54 | 32 | |
| Text-to-Image Generation | OneIG EN | Alignment80.4 | 17 |