Data to 7 Oct 2026

MJ-Bench: Is Your Multimodal Reward Model Really a Good Judge for Text-to-Image Generation?

arXiv 2407.04842 ↗First posted 5 Jul 2024cs.CV cs.CL cs.LGcited by 64

On the map: Diffusion models → Image generation, text-to-image and editing

also in 1 more

LLM alignment / RLHF / preference optimisation → Diffusion and multimodal model alignment

Abstract (arXiv)

While text-to-image models like DALLE-3 and Stable Diffusion are rapidly proliferating, they often encounter challenges such as hallucination, bias, and the production of unsafe, low-quality output. To effectively address these issues, it is crucial to align these models with desired behaviors based on feedback from a multimodal judge. Despite their significance, current multimodal judges frequently undergo inadequate evaluation of their capabilities and limitations, potentially leading to misalignment and unsafe fine-tuning outcomes. To address this issue, we introduce MJ-Bench, a novel benchmark which incorporates a comprehensive preference dataset to evaluate multimodal judges in providing feedback for image generation models across four key perspectives: alignment, safety, image quality, and bias. Specifically, we evaluate a large variety of multimodal judges including smaller-sized CLIP-based scoring models, open-source VLMs (e.g. LLaVA family), and close-source VLMs (e.g. GPT-4o, Claude 3) on each decomposed subcategory of our preference dataset. Experiments reveal that close-source VLMs generally provide better feedback, with GPT-4o outperforming other judges in average. Compared with open-source VLMs, smaller-sized scoring models can provide better feedback regarding text-image alignment and image quality, while VLMs provide more accurate feedback regarding safety and generation bias due to their stronger reasoning capabilities. Further studies in feedback scale reveal that VLM judges can generally provide more accurate and stable feedback in natural language (Likert-scale) than numerical scales. Notably, human evaluations on end-to-end fine-tuned models using separate feedback from these multimodal judges provide similar conclusions, further confirming the effectiveness of MJ-Bench. All data, code, models are available at https://huggingface.co/MJ-Bench.

Key facts. Cited by 64 papers in the map (data through 7 Oct 2026). Placed in Image generation, text-to-image and editing, one of 14 directions in Diffusion models. Builds on 84 of its 97 references that are in the map. First posted on arXiv 5 Jul 2024.

Builds on

84 of 97 references.

Closest ancestors · 7

Same direction, newest first

Other references · 77

2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,456
2020Denoising Diffusion Probabilistic Modelscited by 20,851
2021High-Resolution Image Synthesis with Latent Diffusion Modelscited by 18,694
2023GPT-4 Technical Reportcited by 18,132
2014Microsoft COCO: Common Objects in Contextcited by 17,498
2022Training language models to follow instructions with human feedbackcited by 15,712
2023Llama 2: Open Foundation and Fine-Tuned Chat Modelscited by 12,865
2021Training Verifiers to Solve Math Word Problemscited by 9,743
2023Judging LLM-as-a-Judge with MT-Bench and Chatbot Arenacited by 8,393
2023Direct Preference Optimization: Your Language Model is Secretly a Reward Modelcited by 7,916
2023Visual Instruction Tuningcited by 7,792
2023BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Modelscited by 6,069

38 more

2023Improved Baselines with Visual Instruction Tuningcited by 4,362
2022Flamingo: a Visual Language Model for Few-Shot Learningcited by 4,337
2023SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesiscited by 4,290
2021Zero-Shot Text-to-Image Generationcited by 4,276
2022LAION-5B: An open large-scale dataset for training next generation image-text modelscited by 4,131
2021Scaling Up Visual and Vision-Language Representation Learning With Noisy Text Supervisioncited by 3,848
2017Deep reinforcement learning from human preferencescited by 3,846
2022Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedbackcited by 3,634
2023Grounding DINO: Marrying DINO with Grounded Pre-Training for Open-Set Object Detectioncited by 3,274
2023Tree of Thoughts: Deliberate Problem Solving with Large Language Modelscited by 3,155
2023InstructBLIP: Towards General-purpose Vision-Language Models with Instruction Tuningcited by 2,855
2023MiniGPT-4: Enhancing Vision-Language Understanding with Advanced Large Language Modelscited by 2,689
2020Learning to summarize from human feedbackcited by 2,562
2016Making the V in VQA Matter: Elevating the Role of Image Understanding in Visual Question Answeringcited by 2,498
2021CLIPScore: A Reference-free Evaluation Metric for Image Captioningcited by 2,388
2023Qwen-VL: A Versatile Vision-Language Model for Understanding, Localization, Text Reading, and Beyondcited by 2,179
2019Fine-Tuning Language Models from Human Preferencescited by 2,124
2023MMMU: A Massive Multi-discipline Multimodal Understanding and Reasoning Benchmark for Expert AGIcited by 2,044
2021Align before Fuse: Vision and Language Representation Learning with Momentum Distillationcited by 1,612
2024How Far Are We to GPT-4V? Closing the Gap to Commercial Multimodal Models with Open-Source Suitescited by 1,033
2023ImageReward: Learning and Evaluating Human Preferences for Text-to-Image Generationcited by 1,010
2023Pick-a-Pic: An Open Dataset of User Preferences for Text-to-Image Generationcited by 950
2023Training Diffusion Models with Reinforcement Learningcited by 939
2022Solving math word problems with process- and outcome-based feedbackcited by 871
2023Human Preference Score v2: A Solid Benchmark for Evaluating Human Preferences of Text-to-Image Synthesiscited by 847
2023Diffusion Model Alignment Using Direct Preference Optimizationcited by 767
2023Aligning Large Multimodal Models with Factually Augmented RLHFcited by 703
2018Object Hallucination in Image Captioningcited by 661
2023LLM-Blender: Ensembling Large Language Models with Pairwise Ranking and Generative Fusioncited by 550
2023DecodingTrust: A Comprehensive Assessment of Trustworthiness in GPT Modelscited by 505
2022Safe Latent Diffusion: Mitigating Inappropriate Degeneration in Diffusion Modelscited by 501
2023Scaling up GANs for Text-to-Image Synthesiscited by 492
2024RewardBench: Evaluating Reward Models for Language Modelingcited by 480
2023OPERA: Alleviating Hallucination in Multi-Modal Large Language Models via Over-Trust Penalty and Retrospection-Allocationcited by 468
2021VATT: Transformers for Multimodal Self-Supervised Learning from Raw Video, Audio and Textcited by 427
2021Dynamic Neural Networks: A Surveycited by 405
2024MLLM-as-a-Judge: Assessing Multimodal LLM-as-a-Judge with Vision-Language Benchmarkcited by 362
2021How Much Can CLIP Benefit Vision-and-Language Tasks?cited by 321

50 most cited of 77 All 77 (JSON)

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2024: 18; 2025: 33; 2026: 13.

64 papers.

Sort:

By area:

2025Inference-Time Scaling for Diffusion Models beyond Scaling Denoising Stepscited by 225
2024Generative Reward Modelscited by 148
2024VisionReward: Fine-Grained Multi-Dimensional Human Preference Learning for Image and Video Generationcited by 148
2025RewardBench 2: Advancing Reward Model Evaluationcited by 136
2024GRAPE: Generalizing Robot Policy via Preference Alignmentcited by 106
2024Multi-IF: Benchmarking LLMs on Multi-Turn and Multilingual Instructions Followingcited by 94
2024VL-RewardBench: A Challenging Benchmark for Vision-Language Generative Reward Modelscited by 85
2024MMed-RAG: Versatile Multimodal RAG System for Medical Vision Language Modelscited by 84
2025LEGION: Learning to Ground and Explain for Synthetic Image Detectioncited by 65
2025Video-Bench: Human-Aligned Video Generation Benchmarkcited by 53
2025A Comprehensive Survey of Reward Models: Taxonomy, Applications, Challenges, and Futurecited by 53
2025Reinforcement Learning from Human Feedbackcited by 53

38 more

2025Beyond Reward Hacking: Causal Rewards for Large Language Model Alignmentcited by 42
2024SafeWatch: An Efficient Safety-Policy Following Video Guardrail Model with Transparent Explanationscited by 37
2025The Devil behind the mask: An emergent safety vulnerability of Diffusion LLMscited by 32
2025CAFe: Unifying Representation and Generation with Contrastive-Autoregressive Finetuningcited by 31
2025Shifting AI Efficiency From Model-Centric to Data-Centric Compressioncited by 29
2024Is Your World Simulator a Good Story Presenter? A Consecutive Events-Based Benchmark for Future Long Video Generationcited by 28
2025BlenderGym: Benchmarking Foundational Model Systems for Graphics Editingcited by 27
2025Personalized Preference Fine-tuning of Diffusion Modelscited by 22
2025Efficient Multi-modal Large Language Models via Progressive Consistency Distillationcited by 22
2025AudioMarathon: A Comprehensive Benchmark for Long-Context Audio Understanding and Efficiency in Audio LLMscited by 19
2024Quantifying Generalization Complexity for Large Language Modelscited by 18
2024Img-Diff: Contrastive Data Synthesis for Multimodal Large Language Modelscited by 17
2024Fine-Grained Verifiers: Preference Modeling as Next-token Prediction in Vision-Language Alignmentcited by 17
2024Preference Optimization with Multi-Sample Comparisonscited by 16
2025Prune2Drive: A Plug-and-Play Framework for Accelerating Vision-Language Models in Autonomous Drivingcited by 15
2025MJ-VIDEO: Fine-Grained Benchmarking and Rewarding Video Preferences in Video Generationcited by 14
2025Aligning Multimodal LLM with Human Preference: A Surveycited by 12
2024FactCheckmate: Preemptively Detecting and Mitigating Hallucinations in LMscited by 11
2025Omni-Reward: Towards Generalist Omni-Modal Reward Modeling with Free-Form Preferencescited by 11
2025Benchmarking Multimodal CoT Reward Model Stepwise by Visual Programcited by 10
2025Anyprefer: An Agentic Framework for Preference Data Synthesiscited by 10
2024Safe + Safe = Unsafe? Exploring How Safe Images Can Be Exploited to Jailbreak Large Vision-Language Modelscited by 9
2025Cer-Eval: Certifiable and Cost-Efficient Evaluation Framework for LLMscited by 9
2025Multimodal LLMs as Customized Reward Models for Text-to-Image Generationcited by 9
2026Scientific Image Synthesis: Benchmarking, Methodologies, and Downstream Utilitycited by 8
2025UPME: An Unsupervised Peer Review Framework for Multimodal Large Language Model Evaluationcited by 7
2025Enhancing Large Language Model Reasoning with Reward Models: An Analytical Surveycited by 7
2026Innovator-VL: A Multimodal Large Language Model for Scientific Discoverycited by 7
2025Flex-Judge: Text-Only Reasoning Unleashes Zero-Shot Multimodal Evaluatorscited by 6
2025KRETA: A Benchmark for Korean Reading and Reasoning in Text-Rich VQA Attuned to Diverse Visual Contextscited by 6
2025From EduVisBench to EduVisAgent: A Benchmark and Multi-Agent Framework for Reasoning-Driven Pedagogical Visualizationcited by 5
2025IPCV: Information-Preserving Compression for MLLM Visual Encoderscited by 5
2026Reward Modeling for Reinforcement Learning-Based LLM Reasoning: Design, Challenges, and Evaluationcited by 5
2024Evaluating Semantic Variation in Text-to-Image Synthesis: A Causal Perspectivecited by 4
2024AIDBench: A benchmark for evaluating the authorship identification capability of large language modelscited by 4
2025Libra: Assessing and Improving Reward Model by Learning to Thinkcited by 4
2025RoleRMBench & RoleRM: Towards Reward Modeling for Profile-Based Role Play in Dialogue Systemscited by 4
2025S'MoRE: Structural Mixture of Residual Experts for Parameter-Efficient LLM Fine-tuningcited by 3

50 most cited of 64 All 64 (JSON)

Similar, not cited

Close in wording, no citation either way