Beyond Patches: Global-aware Autoregressive Font Generation
A global-aware autoregressive framework for multimodal few-shot font generation, enabling coherent structure and flexible language-guided style control.
Portfolio
MMMG: A Massive, Multidisciplinary, Multi-Tier Generation Benchmark for Text-to-Image Reasoning
A massive, multidisciplinary benchmark for evaluating multimodal reasoning in knowledge image generation across disciplines, educational levels, and visual formats.