Apple researchers unveil new task to test visual concept inference in AI models

Apple Machine Learning Research has introduced a new task called Visual Concept Inference from Sets (VICIS) to evaluate the ability of vision-language models to infer shared concepts from image sets. Current state-of-the-art models struggle with this, often failing to generalize or defaulting to biased outputs. The researchers propose a novel training framework and architecture designed to infer visual concepts from image sets and extract concept-specific embeddings, demonstrating improved accuracy and diversity on synthetic and large-scale datasets. AI

IMPACT This research aims to improve AI's ability to understand and generate images based on visual context, potentially enhancing multimodal AI capabilities.

RANK_REASON The item describes a new research paper and task proposed by Apple's machine learning research division. [lever_c_demoted from research: ic=1 ai=1.0]

Read on Apple Machine Learning Research →

• Apple Inc.
• Björn Ommer
• ImageNet
• Josh Susskind
• Kolja Bauer
• Ludwig-Maximilians-Universität München
• Miguel Angel Bautista Martin
• Nick Stracke
• Transformer Models
• vision-language models
• Visual Concept Inference from Sets (VICIS)
• WordNet

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →