LESSWRONG
LW

2418
goldemerald
3010
Message
Dialogue
Subscribe

Posts

Sorted by New

Wikitag Contributions

Comments

Sorted by
Newest
No posts to display.
No wikitag contributions to display.
Are SAE features from the Base Model still meaningful to LLaVA?
goldemerald9mo40

As one of the author's of llava-gemma, I'm thrilled to see our model is useful for alignment research. I think the multi-modal aspect of SAE research is under studied, and I am very glad to see this. I will be keeping an eye out for future developments!

Reply