Can I read Starflow2: Bridging Language Models and Normalizing Flows For Unified Multimodal Generation on EtoBox?

Starflow2: Bridging Language Models and Normalizing Flows For Unified Multimodal Generation by shiyili778 is a document available to read on EtoBox.

What is Starflow2: Bridging Language Models and Normalizing Flows For Unified Multimodal Generation about?

STARFlow2 is a unified multimodal model that integrates language models and normalizing flows for generating interleaved text-image sequences without sacrificing visual fidelity or pretrained understanding. It employs the Pretzel architecture, which interleaves a frozen pretrained vision-language model (VLM) with a TARFlow stream, allowing for high-fidelity continuous image generation and structural unification under a single causal mechanism. Experimental results demonstrate STARFlow2

Author

shiyili778

Language

EN