I wonder how llama affects the simulation. Even llama2 consistently fails simple reasoning tasks. (Though I suppose people aren't very reasonable, so it might

I wonder how llama affects the simulation. Even llama2 consistently fails simple reasoning tasks. (Though I suppose people aren't very reasonable, so it might give an accurate simulation!)