Built independently by an author, for readers. Read the story and support ChapterPal
ChapterPal Logo
Log inSign up

keyword

image-text alignment scores

Image-text alignment scores are quantitative metrics used to evaluate how accurately and faithfully the visual content of an image corresponds to the semantic meaning of an accompanying text description or prompt. In multimodal artificial intelligence and text-to-image synthesis, these scores assess whether key elements specified in the text, such as objects, attributes, actions, and spatial relationships, are correctly depicted in the visual output. They are commonly determined through human judgment protocols or automated scoring models, such as vision-language networks that calculate the cosine similarity between text and image embeddings in a shared representation space. These scores provide a standardized way to measure semantic fidelity independently of general visual quality metrics, helping researchers evaluate and compare how well generative models follow natural language instructions.

1 item

Scaling Autoregressive Models for Content-Rich Text-to-Image Generation

Scaling Autoregressive Models for Content-Rich Text-to-Image Generation

Jiahui Yu, Yuanzhong Xu, Jing Yu Koh, Thang Luong, Gunjan Baid, Zirui Wang, Vijay Vasudevan, Alexander Ku, Yinfei Yang, Burcu Karagol Ayan, Ben Hutchinson, Wei Han, Zarana Parekh, Xin Li, Han Zhang, Jason Baldridge, Yonghui Wu

Why you should read this

Demonstrates that scaling an autoregressive sequence-to-sequence model up to 20 billion parameters achieves state-of-the-art text-to-image synthesis quality and effectively handles complex, compositionally rich prompts.

We present the Pathways Autoregressive Text-to-Image (Parti) model, which generates high-fidelity photorealistic images and supports content-rich synthesis involving complex compositions and world knowledge. Parti treats text-to-image generation as a sequence-to-sequence modeling problem, akin to machine translation, with sequences of image tokens as the target outputs rather than text tokens in another language. This strategy can naturally tap into the rich body of prior work on large language models, which have seen continued advances in capabilities and performance through scaling data and model sizes. Our approach is simple: First, Parti uses a Transformer-based image tokenizer, ViT-VQGAN, to encode images as sequences of discrete tokens. Second, we achieve consistent quality improvements by scaling the encoder-decoder Transformer model up to 20B parameters, with a new state-of-the-art zero-shot FID score of 7.23 and finetuned FID score of 3.22 on MS-COCO. Our detailed analysis on Localized Narratives as well as PartiPrompts (P2), a new holistic benchmark of over 1600 English prompts, demonstrate the effectiveness of Parti across a wide variety of categories and difficulty aspects. We also explore and highlight limitations of our models in order to define and exemplify key areas of focus for further improvements. See this https URL for high-resolution images.

Added

2026-09-24