Open stefan-apollo opened 6 months ago
Specifically embedding layer in seed0 model. Seems to be just by 1 though, so lower confidence that something is terribly wrong
Let's see if this is going to be fixed by fixing the Lambdas (and using the new basis), via #244
See OP report results here![image](https://github.com/ApolloResearch/rib/assets/148209923/da3c4084-495e-4a08-8899-7f7615b1784b)