DETONATE: A Benchmark for Text-to-Image Alignment and Kernelized Direct Preference Optimization

TL;DR

This paper introduces DPO-Kernels, an advanced alignment method for text-to-image models, and presents DETONATE, a large-scale benchmark dataset to evaluate safety and bias in T2I systems, along with a new alignment quality metric.

Contribution

It extends Direct Preference Optimization with kernelized representations and divergence choices, and provides the first comprehensive benchmark for T2I alignment and safety evaluation.

Findings

DPO-Kernels improve alignment and safety in T2I models.

DETONATE dataset enables large-scale bias and safety assessment.

Alignment Quality Index reveals hidden vulnerabilities in T2I models.

Abstract

Alignment is crucial for text-to-image (T2I) models to ensure that generated images faithfully capture user intent while maintaining safety and fairness. Direct Preference Optimization (DPO), prominent in large language models (LLMs), is extending its influence to T2I systems. This paper introduces DPO-Kernels for T2I models, a novel extension enhancing alignment across three dimensions: (i) Hybrid Loss, integrating embedding-based objectives with traditional probability-based loss for improved optimization; (ii) Kernelized Representations, employing Radial Basis Function (RBF), Polynomial, and Wavelet kernels for richer feature transformations and better separation between safe and unsafe inputs; and (iii) Divergence Selection, expanding beyond DPO's default Kullback-Leibler (KL) regularizer by incorporating Wasserstein and R'enyi divergences for enhanced stability and robustness. We introduce DETONATE, the first large-scale benchmark of its kind, comprising approximately 100K curated image pairs categorized as chosen and rejected. DETONATE encapsulates three axes of social bias and discrimination: Race, Gender, and Disability. Prompts are sourced from hate speech datasets, with images generated by leading T2I models including Stable Diffusion 3.5 Large, Stable Diffusion XL, and Midjourney. Additionally, we propose the Alignment Quality Index (AQI), a novel geometric measure quantifying latent-space separability of safe/unsafe image activations, revealing hidden vulnerabilities. Empirically, we demonstrate that DPO-Kernels maintain strong generalization bounds via Heavy-Tailed Self-Regularization (HT-SR). DETONATE and complete code are publicly released.