Curated dataset of 1,000 captioned nuclear energy images with automated extraction pipeline and released SDXL checkpoint

Product or Programme · All products and programmes
Evidence · 5 claims (5 for) · 1 study (1 design), q2 · 0 of 1 report an effect size · 5 claims rest on one study

Description

Mohammed I. Radaideh et al. released a curated, captioned nuclear-energy image dataset with preprocessing code, an SDXL checkpoint, and testing prompts for generating technical visual representations.

Components

Claims

• Human qualitative assessment shows SDXL significantly outperforms SD-v3.5-Medium after fine-tuning, failing on 84 versus 250 of 300 prompts [+W]
• Fine-tuning on the 1,000-image nuclear dataset improves image fidelity of SD models, with KID and CMMD decreasing over training steps [+W]
• Commercial text-to-image models produce accurate images for broad nuclear concepts but fail on specialized engineering prompts [+W]
• Generating multiple images per prompt and re-prompting with word variations did not yield accurate nuclear images for failed concepts [+W]
• The CLIP text-image alignment score failed to capture fine-tuning improvements, contradicting KID, CMMD, and human judgment [+W]

Related Products and Programmes

-

Key Sources

• Mohammed I. Radaideh, Jeremy Moon, Andre Gala-Garza, Emma Son, Yug Shah, Majdi I. Radaideh. (2026). NuclearDiffusion: Text-to-Image Foundation Models for Learning Nuclear Energy Concepts. Preprint submitted to Peer Review. https://arxiv.org/abs/2608.04030

Page metadata

Field	Value
type	product
id	curated-dataset-of-1000-captioned-nuclear-energy-images-with-automated-extraction-pipeline
title	Curated dataset of 1,000 captioned nuclear energy images with automated extraction pipeline and released SDXL checkpoint
description	Mohammed I.
status	review
product_kind	dataset
generated	by process:settle-candidates on 2026-10-11
trust	unverified — no one has confirmed this page's content

Frontmatter is the page's machine-readable record — see the Schema & Guide.