mlsanityData to 7 Oct 2026

Embedding an Ethical Mind: Aligning Text-to-Image Synthesis via Lightweight Value Optimization

arXiv 2410.12700 ↗First posted 16 Oct 2024cs.CV cs.AI cs.CYcited by 2

On the map: Diffusion models → Diffusion and flow policies for robotics

also in 2 more
Abstract (arXiv)

Recent advancements in diffusion models trained on large-scale data have enabled the generation of indistinguishable human-level images, yet they often produce harmful content misaligned with human values, e.g., social bias, and offensive content. Despite extensive research on Large Language Models (LLMs), the challenge of Text-to-Image (T2I) model alignment remains largely unexplored. Addressing this problem, we propose LiVO (Lightweight Value Optimization), a novel lightweight method for aligning T2I models with human values. LiVO only optimizes a plug-and-play value encoder to integrate a specified value principle with the input prompt, allowing the control of generated images over both semantics and values. Specifically, we design a diffusion model-tailored preference optimization loss, which theoretically approximates the Bradley-Terry model used in LLM alignment but provides a more flexible trade-off between image quality and value conformity. To optimize the value encoder, we also develop a framework to automatically construct a text-image preference dataset of 86k (prompt, aligned image, violating image, value principle) samples. Without updating most model parameters and through adaptive value selection from the input prompt, LiVO significantly reduces harmful outputs and achieves faster convergence, surpassing several strong baselines and taking an initial step towards ethically aligned T2I models.

Key facts. Cited by 2 papers in the map (data through 7 Oct 2026). Placed in Diffusion and flow policies for robotics, one of 14 directions in Diffusion models. Builds on 46 of its 73 references that are in the map. First posted on arXiv 16 Oct 2024.

Builds on

46 of 73 references.

References · 46

  1. 2022Training language models to follow instructions with human feedbackcited by 15,712
  2. 2019Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformercited by 14,491
  3. 2023LLaMA: Open and Efficient Foundation Language Modelscited by 13,958
  4. 2020Denoising Diffusion Implicit Modelscited by 9,655
  5. 2023Direct Preference Optimization: Your Language Model is Secretly a Reward Modelcited by 7,916
  6. 2023Visual Instruction Tuningcited by 7,792
  7. 2022Hierarchical Text-Conditional Image Generation with CLIP Latentscited by 6,539
  8. 2022Photorealistic Text-to-Image Diffusion Models with Deep Language Understandingcited by 6,322
  9. 2022LAION-5B: An open large-scale dataset for training next generation image-text modelscited by 4,131
  10. 2021On the Opportunities and Risks of Foundation Modelscited by 3,961
  11. 2021The Power of Scale for Parameter-Efficient Prompt Tuningcited by 3,769
  12. 2021Improved Denoising Diffusion Probabilistic Modelscited by 3,736
34 more
  1. 2022Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedbackcited by 3,634
  2. 2016Improved Techniques for Training GANscited by 2,938
  3. 2022An Image is Worth One Word: Personalizing Text-to-Image Generation using Textual Inversioncited by 2,231
  4. 2022DPM-Solver: A Fast ODE Solver for Diffusion Probabilistic Model Sampling in Around 10 Stepscited by 1,961
  5. 2020RealToxicityPrompts: Evaluating Neural Toxic Degeneration in Language Modelscited by 1,235
  6. 2023Kosmos-2: Grounding Multimodal Large Language Models to the Worldcited by 1,075
  7. 2023A General Theoretical Paradigm to Understand Learning from Human Preferencescited by 906
  8. 2020Approximate Nearest Neighbor Negative Contrastive Learning for Dense Text Retrievalcited by 846
  9. 2017Natural TTS Synthesis by Conditioning WaveNet on Mel Spectrogram Predictionscited by 800
  10. 2016Generative Adversarial Text to Image Synthesiscited by 735
  11. 2020FastSpeech 2: Fast and High-Quality End-to-End Text to Speechcited by 710
  12. 2016StackGAN: Text to Photo-realistic Image Synthesis with Stacked Generative Adversarial Networkscited by 575
  13. 2023Language Is Not All You Need: Aligning Perception with Language Modelscited by 543
  14. 2022GIT: A Generative Image-to-text Transformer for Vision and Languagecited by 536
  15. 2023Unified Concept Editing in Diffusion Modelscited by 377
  16. 2022Self-critiquing models for assisting human evaluatorscited by 372
  17. 2019DM-GAN: Dynamic Memory Generative Adversarial Networks for Text-to-Image Synthesiscited by 308
  18. 2021Towards Understanding and Mitigating Social Biases in Language Modelscited by 294
  19. 2023Preference Ranking Optimization for Human Alignmentcited by 283
  20. 2021TrOCR: Transformer-based Optical Character Recognition with Pre-trained Modelscited by 278
  21. 2020Glow-TTS: A Generative Flow for Text-to-Speech via Monotonic Alignment Searchcited by 269
  22. 2023The Unlocking Spell on Base LLMs: Rethinking Alignment via In-Context Learningcited by 251
  23. 2023Scaling Autoregressive Multi-Modal Models: Pretraining and Instruction Tuningcited by 168
  24. 2019Towards Fairer Datasets: Filtering and Balancing the Distribution of the People Subtree in the ImageNet Hierarchycited by 156
  25. 2023Debiasing Vision-Language Models via Biased Promptscited by 147
  26. 2023Social Biases through the Text-to-Image Generation Lenscited by 126
  27. 2024Safety of Multimodal Large Language Models on Images and Textscited by 70
  28. 2023Typology of Risks of Generative Text-to-Image Modelscited by 55
  29. 2023Students' Voices on Generative AI: Perceptions, Benefits, and Challenges in Higher Educationcited by 41
  30. 2022MaskOCR: Text Recognition with Masked Encoder-Decoder Pretrainingcited by 34
  31. 2022Unified Detoxifying and Debiasing in Language Generation via Inference-time Adaptive Optimizationcited by 32
  32. 2023ToViLaG: Your Visual-Language Generative Model is Also An Evildoercited by 27
  33. 2023Mitigating stereotypical biases in text to image generative systemscited by 19
  34. 2023Diffusion idea exploration for art generationcited by 1

Built on by

Cited by year

Citing papers per year: 2025: 2.

2 papers.

  1. 2025Minimalist Concept Erasure in Generative Modelscited by 13
  2. 2025MoHoBench: Assessing Honesty of Multimodal Large Language Models via Unanswerable Visual Questionscited by 10

Similar, not cited

Close in wording, no citation either way

  1. 2025Free Lunch Alignment of Text-to-Image Diffusion Models without Preference Image Pairs
  2. 2024Improving Long-Text Alignment for Text-to-Image Diffusion Models
  3. 2025Instant Preference Alignment for Text-to-Image Diffusion Models
  4. 2024Diffusion-RPO: Aligning Diffusion Models through Relative Preference Optimization
  5. 2025InstructEngine: Instruction-driven Text-to-Image Alignment
  6. 2024Aligning Diffusion Models with Noise-Conditioned Perception
  7. 2025BalancedDPO: Adaptive Multi-Metric Alignment
  8. 2024Multimodal Large Language Model is a Human-Aligned Annotator for Text-to-Image Generation