Training method aligns image understanding and generation in AI models

Original: Semantic Generative Tuning for Unified Multimodal Models

Source: arxiv.org ↗