Announcement
RLCR recipe is being used to train Avalon-M1 models, starting with a 120B model
A researcher on the eight-person team says the RLCR recipe is producing interesting results at a small scale.
TLDR
A researcher says their eight-person team is using its RLCR recipe to train the Avalon-M1 series, starting with a small 120B model. They describe it as early work on agent collectives, with experiments and training still underway.
Combined views
4.5K
25 Sources, first seen ago
