RESEARCH
A cognitive layer for the Moon
Lunarlab in 2026 will be developing a cognitive layer for the lunar frontier. Building on the release of the first multimodal lunar foundation model, Lunar-FM, an explosion of capabilities have been unlocked.
Creating a next generation multimodal multiresolution lunar foundation model for the lunar poles?
Lunarlab 2026 will exploit high resolution NAC imagery, together with lower resolution products to provide a foundation model with multimodal pixel level embeddings at ~1m. The challenge will incorporate physics-based constraints in DEM reconstruction from shadows, and generate multimodal tokenizers for use with LLMs, enabling a step change in possible downstream operational applications.
LUNAR-FM 2.0 | Lunar Poles
