# On Fairness of Unified Multimodal Large Language Model for Image Generation

arXiv: https://arxiv.org/abs/2502.03429 · HTML: https://mlsanity.com/paper/2502.03429/

## Key facts

- Cited by 6 papers in the map (data through 14 Oct 2026).
- Placed in Social bias and fairness in VLMs, one of 16 directions in Multimodal LLMs / vision-language models.
- Builds on 30 of its 38 references that are in the map.
- First posted on arXiv 5 Feb 2025.

## Placed in

- Multimodal LLMs / vision-language models / [Social bias and fairness in VLMs](https://mlsanity.com/mllm/social-bias-and-fairness-in-vlms/)

## Builds on (most cited first, 30 of 30)

- [LoRA: Low-Rank Adaptation of Large Language Models](https://mlsanity.com/paper/2106.09685/) (2021)
- [Direct Preference Optimization: Your Language Model is Secretly a Reward Model](https://mlsanity.com/paper/2305.18290/) (2023)
- [Visual Instruction Tuning](https://mlsanity.com/paper/2304.08485/) (2023)
- [Neural Discrete Representation Learning](https://mlsanity.com/paper/1711.00937/) (2017)
- [Improved Baselines with Visual Instruction Tuning](https://mlsanity.com/paper/2310.03744/) (2023)
- [Taming Transformers for High-Resolution Image Synthesis](https://mlsanity.com/paper/2012.09841/) (2020)
- [CLIPScore: A Reference-free Evaluation Metric for Image Captioning](https://mlsanity.com/paper/2104.08718/) (2021)
- [LLaVA-Med: Training a Large Language-and-Vision Assistant for Biomedicine in One Day](https://mlsanity.com/paper/2306.00890/) (2023)
- [SimPO: Simple Preference Optimization with a Reference-Free Reward](https://mlsanity.com/paper/2405.14734/) (2024)
- [A General Theoretical Paradigm to Understand Learning from Human Preferences](https://mlsanity.com/paper/2310.12036/) (2023)
- [Emu3: Next-Token Prediction is All You Need](https://mlsanity.com/paper/2409.18869/) (2024)
- [Show-o: One Single Transformer to Unify Multimodal Understanding and Generation](https://mlsanity.com/paper/2408.12528/) (2024)
- [Transfusion: Predict the Next Token and Diffuse Images with One Multi-Modal Model](https://mlsanity.com/paper/2408.11039/) (2024)
- [The Reversal Curse: LLMs trained on "A is B" fail to learn "B is A"](https://mlsanity.com/paper/2309.12288/) (2023)
- [VILA-U: a Unified Foundation Model Integrating Visual Understanding and Generation](https://mlsanity.com/paper/2409.04429/) (2024)
- [A Note on the Inception Score](https://mlsanity.com/paper/1801.01973/) (2018)
- [TokenFlow: Unified Image Tokenizer for Multimodal Understanding and Generation](https://mlsanity.com/paper/2412.03069/) (2024)
- [Gender bias and stereotypes in Large Language Models](https://mlsanity.com/paper/2308.14921/) (2023)
- [Debiasing Vision-Language Models via Biased Prompts](https://mlsanity.com/paper/2302.00070/) (2023)
- [What's In My Big Data?](https://mlsanity.com/paper/2310.20707/) (2023)
- [Direct Preference Optimization with an Offset](https://mlsanity.com/paper/2402.10571/) (2024)
- [How well can Text-to-Image Generative Models understand Ethical Natural Language Interventions?](https://mlsanity.com/paper/2210.15230/) (2022)
- [FairFace: Face Attribute Dataset for Balanced Race, Gender, and Age](https://mlsanity.com/paper/1908.04913/) (2019)
- [Dual Diffusion for Unified Image Generation and Understanding](https://mlsanity.com/paper/2501.00289/) (2024)
- [Concept Algebra for (Score-Based) Text-Controlled Generative Models](https://mlsanity.com/paper/2302.03693/) (2023)
- [4M-21: An Any-to-Any Vision Model for Tens of Tasks and Modalities](https://mlsanity.com/paper/2406.09406/) (2024)
- [CultureLLM: Incorporating Cultural Differences into Large Language Models](https://mlsanity.com/paper/2402.10946/) (2024)
- [Training Unbiased Diffusion Models From Biased Dataset](https://mlsanity.com/paper/2403.01189/) (2024)
- [On Catastrophic Inheritance of Large Foundation Models](https://mlsanity.com/paper/2402.01909/) (2024)
- [From Majority to Minority: A Diffusion-based Augmentation for Underrepresented Groups in Skin Lesion Analysis](https://mlsanity.com/paper/2406.18375/) (2024)

## Built on by (most cited first, 6 of 6)

- [Unified Multimodal Understanding and Generation Models: Advances, Challenges, and Opportunities](https://mlsanity.com/paper/2505.02567/) (2025)
- [Doctor Approved: Generating Medically Accurate Skin Disease Images through AI-Expert Feedback](https://mlsanity.com/paper/2506.12323/) (2025)
- [When Understanding Becomes a Risk: Authenticity and Safety Risks in the Emerging Image Generation Paradigm](https://mlsanity.com/paper/2603.24079/) (2026)
- [Multimodal Large Language Models for Enhanced Traffic Safety: A Comprehensive Review and Future Trends](https://mlsanity.com/paper/2504.16134/) (2025)
- [Is your multimodal large language model a good science tutor?](https://mlsanity.com/paper/2505.06418/) (2025)
- [Bias Analysis in Unconditional Image Generative Models](https://mlsanity.com/paper/2506.09106/) (2025)

## Abstract (arXiv)

Unified multimodal large language models (U-MLLMs) have demonstrated impressive performance in visual understanding and generation in an end-to-end pipeline. Compared with generation-only models (e.g., Stable Diffusion), U-MLLMs may raise new questions about bias in their outputs, which can be affected by their unified capabilities. This gap is particularly concerning given the under-explored risk of propagating harmful stereotypes. In this paper, we benchmark the latest U-MLLMs and find that most exhibit significant demographic biases, such as gender and race bias. To better understand and mitigate this issue, we propose a locate-then-fix strategy, where we audit and show how the individual model component is affected by bias. Our analysis shows that bias originates primarily from the language model. More interestingly, we observe a "partial alignment" phenomenon in U-MLLMs, where understanding bias appears minimal, but generation bias remains substantial. Thus, we propose a novel balanced preference model to balance the demographic distribution with synthetic data. Experiments demonstrate that our approach reduces demographic bias while preserving semantic fidelity. We hope our findings underscore the need for more holistic interpretation and debiasing strategies of U-MLLMs in the future.