Pretrained LLMs that support test-time supervised adaptation by compressing few-shot examples into fast weights, eliminating gradient descent.
Guangsheng Bao
gshbao
·
AI & ML interests
NLP
Recent Activity
submitted a paper 2 days ago
FAAST: Forward-Only Associative Learning via Closed-Form Fast Weights for Test-Time Supervised Adaptation authored a paper 2 days ago
FAAST: Forward-Only Associative Learning via Closed-Form Fast Weights for Test-Time Supervised Adaptation updated a collection 3 days ago
faastOrganizations
