This is a heavily interactive web application, and JavaScript is required. Simple HTML interfaces are possible, but that is not what this is.
Post
David Marx
digthatdata.bsky.social
did:plc:yeplumkwwcfqh5gr7burql3e
Something I've been wondering is how much pre-training investment is optimal as a post-training input. Under the "elicitation" theory, you'd anticipate full pre-training to be optimal, but I wonder if an "under-cooked" checkpoint might be more amenable to finetuning?
2025-03-03T16:40:29.156Z