GASS: Geometry-Aware Spherical Sampling for Disentangled Diversity Enhancement in Text-to-Image Generation

2026-02-19

AI Summary

This work addresses the challenge that existing text-to-image (T2I) generation models struggle to balance semantic alignment with output diversity, often limiting user choice and amplifying societal biases. To overcome this, the authors propose a geometric-aware spherical sampling method that explicitly disentangles prompt-relevant from prompt-irrelevant variation directions in the CLIP embedding space for the first time. By leveraging orthogonal decomposition and geometric projection of CLIP embeddings, the approach extends the generation trajectory along two orthogonal axes to enhance diversity. The method is compatible with various frozen T2I backbones—including U-Net and DiT—and integrates seamlessly into both diffusion and flow-based generative frameworks. Extensive experiments demonstrate that it significantly improves generation diversity across multiple benchmarks and architectures while preserving image fidelity and semantic alignment with minimal degradation.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Natural Language Processing: Generation
• Machine Learning: Deep Generative Models & Autoencoders

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Despite high semantic alignment, modern text-to-image (T2I) generative models still struggle to synthesize diverse images from a given prompt. This lack of diversity not only restricts user choice, but also risks amplifying societal biases. In this work, we enhance the T2I diversity through a geometric lens. Unlike most existing methods that rely primarily on entropy-based guidance to increase sample dissimilarity, we introduce Geometry-Aware Spherical Sampling (GASS) to enhance diversity by explicitly controlling both prompt-dependent and prompt-independent sources of variation. Specifically, we decompose the diversity measure in CLIP embeddings using two orthogonal directions: the text embedding, which captures semantic variation related to the prompt, and an identified orthogonal direction that captures prompt-independent variation (e.g., backgrounds). Based on this decomposition, GASS increases the geometric projection spread of generated image embeddings along both axes and guides the T2I sampling process via expanded predictions along the generation trajectory. Our experiments on different frozen T2I backbones (U-Net and DiT, diffusion and flow) and benchmarks demonstrate the effectiveness of disentangled diversity enhancement with minimal impact on image fidelity and semantic alignment.

Problem

• text-to-image generation
• diversity enhancement
• semantic alignment
• societal bias
• image diversity

Innovation

• Geometry-Aware Spherical Sampling
• Disentangled Diversity
• CLIP Embedding Decomposition
• Text-to-Image Generation
• Orthogonal Variation