FLUX-Reason-6M & PRISM-Bench: A Million-Scale Text-to-Image Reasoning Dataset and Comprehensive Benchmark
Rongyao Fang, Aldrich Yu, Chengqi Duan, Linjiang Huang, Shuai Bai, Yuxuan Cai, Kun Wang, Si Liu, Xihui Liu, Hongsheng Li
Abstract
The advancement of open-source text-to-image (T2I) models has been hindered by the absence of large-scale, reasoning-focused datasets and comprehensive evaluation benchmarks, resulting in a performance gap compared to leading closed-source systems. To address this challenge, We introduce FLUX-Reason-6M and PRISM-Bench (Precise and Robust Image Synthesis Measurement Benchmark). FLUX-Reason-6M is a massive dataset consisting of 6 million high-quality FLUX-generated images and 20 million bilingual (English and Chinese) descriptions specifically designed to teach complex reasoning. The image are organized according to six key characteristics: Imagination, Entity, Text rendering, Style, Affection, and Composition, and design explicit Generation Chain-of-Thought (GCoT) to provide detailed breakdowns of image generation steps. The whole data curation takes 15,000 A100 GPU days, providing the community with a resource previously unattainable outside of large industrial labs. PRISM-Bench offers a novel evaluation standard with seven distinct tracks, including a formidable Long Text challenge using GCoT. Through carefully designed prompts, it utilizes advanced visionlanguage models for nuanced human-aligned assessment of prompt-image alignment and image aesthetics. Our extensive evaluation of 19 leading models on PRISM-Bench reveals critical performance gaps and highlights specific areas requiring improvement. Our dataset, benchmark, and evaluation code are released to catalyze the next wave of reasoning-oriented T2I generation. Composition: Cute Morkie puppy perched inside a pink box labeled "Fabulous". Adorned with a matching pink bow and surrounded by fluffy feathers and vibrant pink flowers, the puppy exudes charm and playfulness. Affection: Enveloped in a world of soft textures and vibrant hues, this Morkie puppy embodies the essence of innocence and wonder, radiating an irresistible charm that ignites a heartwarming sense of joy and hope. GCoT: Cute and charming, a Morkie puppy sits adorably perched inside a pastel pink cardboard box labeled "Fabulous" in bold black script. The puppy, with its fluffy white fur accented by brown patches around its ears and face, exudes an irresistible charm, enhanced by a matching pink satin bow tied neatly atop its head. Its small paws, tipped with dark pads, rest gently on the edge of the box, adding a playful touch to the composition. Surrounding the puppy are fluffy white feathers and vibrant pink flowers, creating a whimsical and enchanting atmosphere. The background is a soft, uniform pink, which complements the overall color scheme and draws attention to the central subject. The image captures a perfect blend of innocence and playfulness, evoking a heartwarming sense of joy and hope through its soft textures, vibrant hues, and meticulous arrangement.
Ask about this paper
Your agent reads all of it.
Lune indexed this paper to the last equation, along with the top-tier papers that cite it. Ask a question and the answer quotes them.
Your agent calls
Luneget_paper_fulltext
Free to start. No credit card required.
Terminal
Install the CLIlune papers fulltext 926fa3fa-c44f-4b62-94a0-0935e7dbc403Cited by top-tier papers13
- Easier Painting Than Thinking: Can Text-to-Image Models Set the Stage, but Not Direct the Play?Ouxiang Li, Yuan Wang, Xinting Hu, Huijuan Huang et al.ICLR 2026 · 39 citations
- NExT-OMNI: Towards Any-to-Any Omnimodal Foundation Models with Discrete Flow MatchingRun Luo, Xiaobo Xia, Lu Wang, Longze Chen et al.ICLR 2026 · 22 citations
- Factuality Matters: When Image Generation and Editing Meet Structured VisualsLe Zhuo, Songhao Han, Yuandong Pu, Boxiang Qiu et al.ICLR 2026 · 15 citations
- TextPecker: Rewarding Structural Anomaly Quantification for Enhancing Visual Text RenderingHanshen Zhu, Yuliang Liu, Xuecheng Wu, An-Lan Wang et al.CVPR 2026 · 15 citations
- ThinkGen: Generalized Thinking for Visual GenerationSiyu Jiao, Yiheng Lin, Yujie Zhong, Qi She et al.CVPR 2026 · 12 citations
Builds on24
- Denoising Diffusion Probabilistic ModelsJonathan Ho, Ajay Jain, Pieter AbbeelNeurIPS 2020 · 35,902 citations
- High-Resolution Image Synthesis with Latent Diffusion ModelsRobin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser et al.CVPR 2022 · 13,123 citations
- Scaling Up Visual and Vision-Language Representation Learning With Noisy Text SupervisionChao Jia, Yinfei Yang, Ye Xia, Yi-Ting Chen et al.ICML 2021 · 5,401 citations
- SDXL: Improving Latent Diffusion Models for High-Resolution Image SynthesisDustin Podell, Zion English, Kyle Lacey, Andreas Blattmann et al.ICLR 2024 · 4,569 citations
- Scaling Rectified Flow Transformers for High-Resolution Image SynthesisPatrick Esser, Sumith Kulal, Andreas Blattmann, Rahim Entezari et al.ICML 2024 · 3,620 citations
Related papers
- GIR-Bench: Versatile Benchmark for Generating Images with ReasoningHongxiang Li, Yaowei Li, Bin Lin, Yuwei Niu et al.ICLR 2026 · 15 citations
- MMTIT-Bench: A Multilingual and Multi-Scenario Benchmark with Cognition-Perception-Reasoning Guided Text-Image Machine TranslationGengluo Li, Chengquan Zhang, Yupu Liang, Huawen Shen et al.CVPR 2026 · 6 citations
- R2I-Bench: Benchmarking Reasoning-Driven Text-to-Image GenerationKaijie Chen, Zihao Lin, Zhiyang Xu, Ying Shen et al.EMNLP 2025
- Will Multimodal Models Be Dazzled by Multi-Image Visual Puzzles?zhi zhu, YaoQi Fan, Zhe Chen, Yue Cao et al.CVPR 2026
- Photorealistic Text-to-Image Diffusion Models with Deep Language UnderstandingChitwan Saharia, William Chan, Saurabh Saxena, Lala Li et al.NeurIPS 2022 · 8,965 citations
