Comment by senordevnyc

I thought the consensus was that models couldn’t actually introspect like this. So there’s no reason to think any of those reasons are actually why the model did what it did, right? Has this changed?