HyperAlign: Hyperbolic Entailment Cones for Adaptive Text-to-Image Alignment Assessment

2026-01-08

arXiv.org

AI Summary

Existing text-to-image alignment evaluation methods predominantly rely on Euclidean space metrics, which struggle to capture semantic hierarchies and lack sample adaptivity. This work proposes the first adaptive alignment evaluation framework grounded in hyperbolic entailment geometry: CLIP features are mapped into hyperbolic space, where hyperbolic entailment cones model semantic inclusion relationships. A dynamic supervision mechanism coupled with an adaptive modulation regressor enables structure-aware, sample-level alignment score prediction. The proposed method achieves state-of-the-art performance in both in-domain evaluation and cross-dataset generalization, demonstrating the efficacy and advantages of hyperbolic geometry for assessing text–image alignment.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Natural Language Processing: Sentence-level Semantics, Textual Inference, etc.
• Machine Learning: Multimodal Learning

Application Category

• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

With the rapid development of text-to-image generation technology, accurately assessing the alignment between generated images and text prompts has become a critical challenge. Existing methods rely on Euclidean space metrics, neglecting the structured nature of semantic alignment, while lacking adaptive capabilities for different samples. To address these limitations, we propose HyperAlign, an adaptive text-to-image alignment assessment framework based on hyperbolic entailment geometry. First, we extract Euclidean features using CLIP and map them to hyperbolic space. Second, we design a dynamic-supervision entailment modeling mechanism that transforms discrete entailment logic into continuous geometric structure supervision. Finally, we propose an adaptive modulation regressor that utilizes hyperbolic geometric features to generate sample-level modulation parameters, adaptively calibrating Euclidean cosine similarity to predict the final score. HyperAlign achieves highly competitive performance on both single database evaluation and cross-database generalization tasks, fully validating the effectiveness of hyperbolic geometric modeling for image-text alignment assessment.

Problem

• text-to-image alignment
• alignment assessment
• semantic structure
• adaptive evaluation
• Euclidean metrics

Innovation

• hyperbolic geometry
• entailment cones
• adaptive alignment assessment
• text-to-image generation
• geometric representation learning