fractale-lm/Fractale-350M-base
Text Generation • Updated • 6
LMs whose long-term memory is a self-written fast-weight thought bank: the model writes its own notes and thinks through them.
Note 386M base model, phase-1 pretraining (release pending)