Machina Mirabilis

An experiment to see if an LLM trained from scratch on text prior to 1900 can come up with quantum mechanics and relativity.

hCA-776

LLM optimized carbon capture enzymes that are 170% more active and 25% more stable, experimentally validated

Pro-1

A reasoning model trained using GRPO towards a physics based reward function for protein stability.

SMILES-MDLM

Masked diffusion language modeling for text guided and chemical property conditioned molecule generation.