Skip to content

MeanFlow Long-term Invariant Spatiotemporal Consistency Autoregressive Models (MeLISA)

Home / Concepts / MeanFlow Long-term Invariant Spatiotemporal Consistency Autoregressive Models (MeLISA)

MeanFlow Long-term Invariant Spatiotemporal Consistency Autoregressive Models (MeLISA)

Auto-generated stub. Edit this file to add more details.

A latent-free autoregressive generative surrogate that uses MeanFlow and consistency-based objectives to achieve efficient, physically accurate long-horizon trajectory forecasting.

Why It Matters

Introduces a novel latent-free autoregressive generative framework that enables single-pass high-resolution temporal forecasting while maintaining long-term physical consistency.

Evidence

We propose MeanFlow Long-term Invariant Spatiotemporal Consistency Autoregressive Models (MeLISA), a latent-free autoregressive generative surrogate built on pixel-space MeanFlow.

Metadata & Links