Embedding an Ethical Mind: Aligning Text-to-Image Synthesis via Lightweight Value Optimization
Open Access
- 28 October 2024
- conference paper
- Published by Association for Computing Machinery (ACM)
- p. 3558-3567
Abstract
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. Warning: This paper involves descriptions and images depicting discriminatory, pornographic, bloody, and horrific scenes.Keywords
All Related Versions
Funding Information
- National Key Research and Development Program of China (2024QY1400)
This publication has 20 references indexed in Scilit:
- Diffusion Model Alignment Using Direct Preference Optimization2024
- Erasing Concepts from Diffusion Models2023
- DreamBooth: Fine Tuning Text-to-Image Diffusion Models for Subject-Driven Generation2023
- Safe Latent Diffusion: Mitigating Inappropriate Degeneration in Diffusion Models2023
- Scene Text Recognition with Permuted Autoregressive Sequence Models2022
- High-Resolution Image Synthesis with Latent Diffusion Models2022
- DM-GAN: Dynamic Memory Generative Adversarial Networks for Text-To-Image Synthesis2019
- AttnGAN: Fine-Grained Text to Image Generation with Attentional Generative Adversarial Networks2018
- Natural TTS Synthesis by Conditioning Wavenet on MEL Spectrogram Predictions2018
- StackGAN: Text to Photo-Realistic Image Synthesis with Stacked Generative Adversarial Networks2017