Reproduce “MUNITE: Unified Multimodal Latent Inference for Any-to-Any Multimodal Generation” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Kyeongmin Yeo, Minhyuk Sung
MUNITE models conditional latent distributions across levels of observed modalities, then uses one sampled latent to generate multiple outputs. Sharing the latent is intended to preserve details left unspecified by the input. On image–text–audio benchmarks, it reports highest coherence in six one-to-many and unconditional comparisons; image-conditioned text-and-audio CLAP similarity is 34.762 versus 32.237 for MUNI. One-to-one and many-to-one performance is competitive.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.