Infinity and Beyond: Compositional Alignment in VAR and Diffusion T2I Models

TL;DR

This paper systematically compares the compositional alignment capabilities of VAR and diffusion-based text-to-image models, revealing strengths and weaknesses across multiple benchmarks and establishing baselines for future research.

Contribution

It provides the first comprehensive benchmark comparison of VAR and diffusion T2I models on compositional tasks, highlighting the strengths of Infinity-8B and Infinity-2B.

Findings

Infinity-8B achieves the best overall compositional alignment.

Infinity-2B matches or exceeds larger diffusion models in several categories.

SDXL and PixArt-α show weaknesses in attribute and spatial tasks.

Abstract

Achieving compositional alignment between textual descriptions and generated images - covering objects, attributes, and spatial relationships - remains a core challenge for modern text-to-image (T2I) models. Although diffusion-based architectures have been widely studied, the compositional behavior of emerging Visual Autoregressive (VAR) models is still largely unexamined. We benchmark six diverse T2I systems - SDXL, PixArt-$\alpha$, Flux-Dev, Flux-Schnell, Infinity-2B, and Infinity-8B - across the full T2I-CompBench++ and GenEval suites, evaluating alignment in color and attribute binding, spatial relations, numeracy, and complex multi-object prompts. Across both benchmarks, Infinity-8B achieves the strongest overall compositional alignment, while Infinity-2B also matches or exceeds larger diffusion models in several categories, highlighting favorable efficiency-performance trade-offs. In contrast, SDXL and PixArt-$\alpha$ show persistent weaknesses in attribute-sensitive and spatial tasks. These results provide the first systematic comparison of VAR and diffusion approaches to compositional alignment and establish unified baselines for the future development of the T2I model.