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

2025-05-01

AI Summary

Weak reasoning capabilities hinder text-to-image (T2I) generation. To address this, we propose BiCoT—a dual-level Chain-of-Thought (CoT) framework synergistically optimized via reinforcement learning: cross-modal planning operates at the semantic level, while pixel-level fine-grained generation is realized at the token level. We introduce the first joint semantic- and token-level CoT modeling mechanism and design BiCoT-GRPO, the first algorithm enabling synchronous optimization of both reasoning pathways within a single inference step. Our approach integrates multi-reward composition and Janus-Pro architecture fine-tuning. Evaluated on T2I-CompBench and WISE benchmarks, BiCoT achieves 13% and 19% improvements, respectively, surpassing state-of-the-art models including FLUX.1. This work establishes a novel, interpretable, and decomposable hierarchical reasoning paradigm for T2I generation.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Natural Language Processing: Generation
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

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

Problem

• Enhancing text-to-image generation with reasoning strategies
• Integrating semantic-level and token-level chain-of-thought processes
• Optimizing generation performance using reinforcement learning and CoT

Innovation

• Uses bi-level CoT for semantic and token reasoning
• Integrates RL with BiCoT-GRPO for coordinated optimization
• Enhances T2I generation via collaborative CoT and RL