Can an agentic pipeline out-prompt bias? A controlled comparison of one-shot vs agentic image generation
Assignment 1: Investigating biases and blind spots in AI systems. All code, prompts, and scoring data are in the public repository; every image on this page is AI-generated, watermarked, and published only to document model behavior. 1. The problem Most products that generate images do it the simplest possible way: the user’s text goes straight […]