/
#3
Nous Research
Open-source AI research group producing the Hermes model family, fine-tuned on LLaMA and Mistral architectures using large-scale synthetic data. Research spans SFT, DPO, and chain-of-thought reasoning. Also developing DisTrO, a distributed pre-training method that cuts inter-GPU bandwidth by orders of magnitude, enabling training across heterogeneous, internet-connected hardware.
Categories
Subcategories
AGENT FRAMEWORKSCOMPUTER USETOOL USE
Links