T2I-R1: Reinforcing Image Generation with Collaborative Semantic-level and Token-level CoT

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Jiang, Dongzhi, Guo, Ziyu, Zhang, Renrui, Zong, Zhuofan, Li, Hao, Zhuo, Le, Yan, Shilin, Heng, Pheng-Ann, Li, Hongsheng
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866909669553864704
author Jiang, Dongzhi
Guo, Ziyu
Zhang, Renrui
Zong, Zhuofan
Li, Hao
Zhuo, Le
Yan, Shilin
Heng, Pheng-Ann
Li, Hongsheng
author_facet Jiang, Dongzhi
Guo, Ziyu
Zhang, Renrui
Zong, Zhuofan
Li, Hao
Zhuo, Le
Yan, Shilin
Heng, Pheng-Ann
Li, Hongsheng
contents 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
format Preprint
id arxiv_https___arxiv_org_abs_2505_00703
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle T2I-R1: Reinforcing Image Generation with Collaborative Semantic-level and Token-level CoT
Jiang, Dongzhi
Guo, Ziyu
Zhang, Renrui
Zong, Zhuofan
Li, Hao
Zhuo, Le
Yan, Shilin
Heng, Pheng-Ann
Li, Hongsheng
Computer Vision and Pattern Recognition
Artificial Intelligence
Computation and Language
Machine Learning
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
title T2I-R1: Reinforcing Image Generation with Collaborative Semantic-level and Token-level CoT
topic Computer Vision and Pattern Recognition
Artificial Intelligence
Computation and Language
Machine Learning
url https://arxiv.org/abs/2505.00703