New framework enables generative models to handle variable dimensions and sequence lengths

Researchers have introduced Expanding Generative Flows (EFlows) and Expanding Flow Maps (EFMs) as a novel framework for generative models. These models can handle variable dimensions and sequence lengths by augmenting the state space with conditional noise. The framework decomposes the generative process into an expand operator, which increases the state space, and a transport map, which advances the state along an interpolant. This approach generalizes existing fixed-canvas flows and extends to discrete domains for variable-size graph and sequence generation. AI

IMPACT This research could lead to more flexible and powerful generative models capable of handling diverse data structures and lengths.

RANK_REASON The cluster contains a research paper detailing a new framework for generative models. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• arXiv
• Expanding Flow Maps
• Expanding Generative Flows
• Hugging Face

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →