AI agents turn photos into editable 3D scenes but struggle to assess accuracy
🤖 AI-generated content — The title and summary were produced automatically by artificial intelligence, without human editorial review.
University of Maryland and AWS researchers developed LEGO-Anything, which has coding agents iteratively reconstruct a single photo as editable Blender code. On LEGO-Bench’s 208 images from 104 scenes, GPT-6 Astra reached up to 53% accuracy, while agents judged their own geometric accuracy no better than chance.
Continue in the app — vote & join in ➔Source: The Decoder (AI) · via ahirlevel.hu