Skip to content

Comment on Self-Adapting Language Modelsparent

Comments

Looks like it may be the stability of the approach, avoiding alignment tax and model collapse.

I'd love to see a full circle of hypernetworks, with both models continuously updated through generated LoRAs, the hypernetwork updated to accommodate the new model state. You'd need a meta-hypernetwork to apply LoRAs to the hypernetwork, and then you could effectively have continuous learning.

AboutSource Built by g1lg1l

Hackerly is an independent reader for Hacker News, built on the public HN API. Not affiliated with Y Combinator.