Research
LARC proposes latent reasoning in continuous space for unified multimodal image generation
The LARC framework uses curriculum SFT and information-gain RL to reason over hidden states instead of tokens. It reports a 7 point GenEval gain over BAGEL and better test-time scaling.
Every outlet that ran it 1 article
Hugging Face PapersResearchLead
Latent Reasoning in Continuous Space for Unified Multimodal Models