mlsanityData to 7 Oct 2026

Can We Generate Images with CoT? Let's Verify and Reinforce Image Generation Step by Step

arXiv 2501.13926 ↗First posted 23 Jan 2025cs.CV cs.AI cs.CLcited by 130

On the map: LLM reasoning → Multimodal reasoning

also in 1 more
Abstract (arXiv)

Chain-of-Thought (CoT) reasoning has been extensively explored in large models to tackle complex understanding tasks. However, it still remains an open question whether such strategies can be applied to verifying and reinforcing image generation scenarios. In this paper, we provide the first comprehensive investigation of the potential of CoT reasoning to enhance autoregressive image generation. We focus on three techniques: scaling test-time computation for verification, aligning model preferences with Direct Preference Optimization (DPO), and integrating these techniques for complementary effects. Our results demonstrate that these approaches can be effectively adapted and combined to significantly improve image generation performance. Furthermore, given the pivotal role of reward models in our findings, we propose the Potential Assessment Reward Model (PARM) and PARM++, specialized for autoregressive image generation. PARM adaptively assesses each generation step through a potential assessment approach, merging the strengths of existing reward models, and PARM++ further introduces a reflection mechanism to self-correct the generated unsatisfactory image, which is the first to incorporate reflection in autoregressive image generation. Using our investigated reasoning strategies, we enhance a baseline model, Show-o, to achieve superior results, with a significant +24% improvement on the GenEval benchmark, surpassing Stable Diffusion 3 by +15%. We hope our study provides unique insights and paves a new path for integrating CoT reasoning with autoregressive image generation. Code and models are released at https://github.com/ZiyuGuo99/Image-Generation-CoT

Key facts. Cited by 130 papers in the map (data through 7 Oct 2026). Placed in Multimodal reasoning, one of 17 directions in LLM reasoning. Builds on 59 of its 92 references that are in the map. First posted on arXiv 23 Jan 2025.

Builds on

59 of 92 references.

Closest ancestors · 6

Same direction, newest first

  1. 2024Improve Vision Language Model Chain-of-thought Reasoning21 Oct
  2. 2024MAVIS: Mathematical Visual Instruction Tuning with an Automatic Data Engine11 Jul
  3. 2023MathVista: Evaluating Mathematical Reasoning of Foundation Models in Visual Contexts3 Oct
  4. 2023MMBench: Is Your Multi-modal Model an All-around Player?12 Jul
  5. 2023LLaMA-Adapter V2: Parameter-Efficient Visual Instruction Model28 Apr
  6. 2023Multimodal Chain-of-Thought Reasoning in Language Models2 Feb

Other references · 53

  1. 2021High-Resolution Image Synthesis with Latent Diffusion Modelscited by 18,694
  2. 2023GPT-4 Technical Reportcited by 18,132
  3. 2017Proximal Policy Optimization Algorithmscited by 14,441
  4. 2023LLaMA: Open and Efficient Foundation Language Modelscited by 13,958
  5. 2023Llama 2: Open Foundation and Fine-Tuned Chat Modelscited by 12,865
  6. 2021Training Verifiers to Solve Math Word Problemscited by 9,743
  7. 2022Hierarchical Text-Conditional Image Generation with CLIP Latentscited by 6,539
  8. 2022Self-Consistency Improves Chain of Thought Reasoning in Language Modelscited by 5,601
  9. 2023SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesiscited by 4,290
  10. 2024Scaling Rectified Flow Transformers for High-Resolution Image Synthesiscited by 4,094
  11. 2023Let's Verify Step by Stepcited by 3,598
  12. 2022Constitutional AI: Harmlessness from AI Feedbackcited by 2,950
38 more
  1. 2024LLaVA-OneVision: Easy Visual Task Transfercited by 2,669
  2. 2024Qwen2 Technical Reportcited by 2,076
  3. 2024Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameterscited by 1,928
  4. 2023MME: A Comprehensive Evaluation Benchmark for Multimodal Large Language Modelscited by 1,624
  5. 2024Video-MME: The First-Ever Comprehensive Evaluation Benchmark of Multi-modal LLMs in Video Analysiscited by 1,375
  6. 2024DeepSeek-Coder: When the Large Language Model Meets Programming -- The Rise of Code Intelligencecited by 1,154
  7. 2023Large Language Models Cannot Self-Correct Reasoning Yetcited by 965
  8. 2024Chameleon: Mixed-Modal Early-Fusion Foundation Modelscited by 904
  9. 2024Emu3: Next-Token Prediction is All You Needcited by 761
  10. 2024Autoregressive Model Beats Diffusion: Llama for Scalable Image Generationcited by 760
  11. 2024Show-o: One Single Transformer to Unify Multimodal Understanding and Generationcited by 734
  12. 2024Autoregressive Image Generation without Vector Quantizationcited by 644
  13. 2023WizardMath: Empowering Mathematical Reasoning for Large Language Models via Reinforced Evol-Instructcited by 618
  14. 2024Transfusion: Predict the Next Token and Diffuse Images with One Multi-Modal Modelcited by 468
  15. 2024Training Language Models to Self-Correct via Reinforcement Learningcited by 405
  16. 2023SLiC-HF: Sequence Likelihood Calibration with Human Feedbackcited by 395
  17. 2024AFlow: Automating Agentic Workflow Generationcited by 395
  18. 2024ReST-MCTS*: LLM Self-Training via Process Reward Guided Tree Searchcited by 338
  19. 2024SEED-X: Multimodal Models with Unified Multi-granularity Comprehension and Generationcited by 328
  20. 2024DeepSeek-Coder-V2: Breaking the Barrier of Closed-Source Models in Code Intelligencecited by 325
  21. 2024Step-DPO: Step-wise Preference Optimization for Long-chain Reasoning of LLMscited by 265
  22. 2024Is DPO Superior to PPO for LLM Alignment? A Comprehensive Studycited by 262
  23. 2024Iterative Reasoning Preference Optimizationcited by 244
  24. 2024Chain-of-Thought Reasoning Without Promptingcited by 208
  25. 2024World Model on Million-Length Video And Language With Blockwise RingAttentioncited by 196
  26. 2023Point-Bind & Point-LLM: Aligning Point Cloud with Multi-modality for 3D Understanding, Generation, and Instruction Followingcited by 192
  27. 2024DeepSeek-Prover-V1.5: Harnessing Proof Assistant Feedback for Reinforcement Learning and Monte-Carlo Tree Searchcited by 189
  28. 2024Teaching Large Language Models to Reason with Reinforcement Learningcited by 158
  29. 2023ImageBind-LLM: Multi-modality Instruction Tuningcited by 142
  30. 2021Dueling RL: Reinforcement Learning with Trajectory Preferencescited by 105
  31. 2024MMSearch: Benchmarking the Potential of Large Models as Multi-modal Search Enginescited by 98
  32. 2023Let's reward step by step: Step-Level reward model as the Navigators for Reasoningcited by 96
  33. 2024CoMat: Aligning Text-to-Image Diffusion Model with Image-to-Text Concept Matchingcited by 60
  34. 2024Make Every Move Count: LLM-based High-Quality RTL Code Generation Using MCTScited by 42
  35. 2024Step-Controlled DPO: Leveraging Stepwise Error for Enhanced Mathematical Reasoningcited by 35
  36. 2024ComfyBench: Benchmarking LLM-based Agents in ComfyUI for Autonomously Designing Collaborative AI Systemscited by 30
  37. 2024SAM2Point: Segment Any 3D as Videos in Zero-shot and Promptable Mannerscited by 18
  38. 2024Code-Optimise: Self-Generated Preference Data for Correctness and Efficiencycited by 16

50 most cited of 53 All 53 (JSON)

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 94; 2026: 36.

130 papers.

Sort:

By area:

  1. 2025DanceGRPO: Unleashing GRPO on Visual Generationcited by 388
  2. 2025Towards Reasoning Era: A Survey of Long Chain-of-Thought for Reasoning Large Language Modelscited by 381
  3. 2025Thinking with Images for Multimodal Reasoning: Foundations, Methods, and Future Frontierscited by 193
  4. 2025HybridVLA: Collaborative Diffusion and Autoregression in a Unified Vision-Language-Action Modelcited by 176
  5. 2025Multimodal Chain-of-Thought Reasoning: A Comprehensive Surveycited by 176
  6. 2025T2I-R1: Reinforcing Image Generation with Collaborative Semantic-level and Token-level CoTcited by 162
  7. 2025MME-CoT: Benchmarking Chain-of-Thought in Large Multimodal Models for Reasoning Quality, Robustness, and Efficiencycited by 123
  8. 2025Echo-4o: Harnessing the Power of GPT-4o Synthetic Images for Improved Image Generationcited by 95
  9. 2025Visual Planning: Let's Think Only with Imagescited by 85
  10. 2025Thinking with Generated Imagescited by 71
  11. 2025TIIF-Bench: How Does Your T2I Model Follow Your Instructions?cited by 64
  12. 2025UniCTokens: Boosting Personalized Understanding and Generation via Unified Concept Tokenscited by 55
38 more
  1. 2025From Reflection to Perfection: Scaling Inference-Time Optimization for Text-to-Image Diffusion Models via Reflection Tuningcited by 53
  2. 2025Uni-cot: Towards Unified Chain-of-Thought Reasoning Across Text and Visioncited by 52
  3. 2025Reconstruction Alignment Improves Unified Multimodal Modelscited by 48
  4. 2025MINT-CoT: Enabling Interleaved Visual Tokens in Mathematical Chain-of-Thought Reasoningcited by 45
  5. 2025Reflect-DiT: Inference-Time Scaling for Text-to-Image Diffusion Transformers via In-Context Reflectioncited by 42
  6. 2025Are Video Models Ready as Zero-Shot Reasoners? An Empirical Study with the MME-CoF Benchmarkcited by 41
  7. 2025Video-T1: Test-Time Scaling for Video Generationcited by 39
  8. 2025Lumina-mGPT 2.0: Stand-Alone AutoRegressive Image Modelingcited by 35
  9. 2025Easier Painting Than Thinking: Can Text-to-Image Models Set the Stage, but Not Direct the Play?cited by 33
  10. 2025RealUnify: Do Unified Models Truly Benefit from Unification? A Comprehensive Benchmarkcited by 33
  11. 2025Detect Anything 3D in the Wildcited by 32
  12. 2025UniGen: Enhanced Training & Test-Time Strategies for Unified Multimodal Understanding and Generationcited by 32
  13. 2025Delving into RL for Image Generation with CoT: A Study on DPO vs. GRPOcited by 31
  14. 2025R2I-Bench: Benchmarking Reasoning-Driven Text-to-Image Generationcited by 31
  15. 2025MindOmni: Unleashing Reasoning Generation in Vision Language Models with RGPOcited by 30
  16. 2025ReasonGen-R1: CoT for Autoregressive Image generation models through SFT and RLcited by 30
  17. 2025Direct Numerical Layout Generation for 3D Indoor Scene Synthesis via Spatial Reasoningcited by 27
  18. 2025textttComplex-Edit: CoT-Like Instruction Generation for Complexity-Controllable Image Editing Benchmarkcited by 25
  19. 2025Why Reasoning Matters? A Survey of Advancements in Multimodal Reasoning (v1)cited by 23
  20. 2025UniRL: Self-Improving Unified Multimodal Models via Supervised and Reinforcement Learningcited by 23
  21. 2025SciVerse: Unveiling the Knowledge Comprehension and Visual Reasoning of LMMs on Multi-modal Scientific Problemscited by 22
  22. 2025RePrompt: Reasoning-Augmented Reprompting for Text-to-Image Generation via Reinforcement Learningcited by 22
  23. 2025PromptSculptor: Multi-Agent Based Text-to-Image Prompt Optimizationcited by 19
  24. 2025A Survey on Agentic Multimodal Large Language Modelscited by 19
  25. 2025SRUM: Fine-Grained Self-Rewarding for Unified Multimodal Modelscited by 19
  26. 2026Visual Generation Unlocks Human-Like Reasoning through Multimodal World Modelscited by 18
  27. 2025Thinking-while-Generating: Interleaving Textual Reasoning throughout Visual Generationcited by 17
  28. 2025Exploring the Potential of Encoder-free Architectures in 3D LMMscited by 16
  29. 2025Group Critical-token Policy Optimization for Autoregressive Image Generationcited by 16
  30. 2025TTS-VAR: A Test-Time Scaling Framework for Visual Auto-Regressive Generationcited by 15
  31. 2025STAGE: Stable and Generalizable GRPO for Autoregressive Image Generationcited by 15
  32. 2025Towards Enhanced Image Generation Via Multi-modal Chain of Thought in Unified Generative Modelscited by 14
  33. 2025Generative AI Act II: Test Time Scaling Drives Cognition Engineeringcited by 14
  34. 2025Sailing by the Stars: A Survey on Reward Models and Learning Strategies for Learning from Rewardscited by 14
  35. 2025Q-Ponder: A Unified Training Pipeline for Reasoning-based Visual Quality Assessmentcited by 14
  36. 2025Reinforcement Fine-Tuning Powers Reasoning Capability of Multimodal Large Language Modelscited by 13
  37. 2025FocusDiff: Advancing Fine-Grained Text-Image Alignment for Autoregressive Visual Generation through RLcited by 13
  38. 2025DraCo: Draft as CoT for Text-to-Image Preview and Rare Concept Generationcited by 13

50 most cited of 130 All 130 (JSON)

Similar, not cited

Close in wording, no citation either way

  1. 2025Visual-Aware CoT: Achieving High-Fidelity Visual Consistency in Unified Models
  2. 2026RL-RIG: A Generative Spatial Reasoner via Intrinsic Reflection
  3. 2026AutoRef: Harness Optimization for Agentic Multi-Reference Image Generation
  4. 2025GoT-R1: Unleashing Reasoning Capability of MLLM for Visual Generation with Reinforcement Learning
  5. 2026Thinking in Pictures: A Systematic Benchmark for Reasoning-driven Image Generation
  6. 2026UniGRPO: Unified Policy Optimization for Reasoning-Driven Visual Generation
  7. 2026Think Before You Score: Thinking Reward Model for Visual Generation
  8. 2026GenPuzzle: Benchmarking Visual Reasoning in Image Generation Models