Shi-Zhe Chen

HyLaR framework enhances multimodal LLMs with hybrid latent reasoning

Researchers have introduced HyLaR, a novel framework designed to enhance multimodal large language models (MLLMs) by integrating discrete text generation with continuous visual latent representations. This approach aims…