Halfway through my time at Austin, I’m actually starting to get used to the blistering heat :)

I got LLAMA-2 generations. The funny thing is that it’s quite conversational Patterns I’ve noticed with LLAMA2 conversational: Sometimes starts out with a rhetorical question: “So, you know how sometimes you feel sad or angry and you don’t know why?” Uses “so,” “we”, and “!” a lot Sometimes randomly puts in Russian words Sometimes ends with followup question(s): “How do you think your mom or dad would answer the questions on the scenarios the actors acted out? Do you think they would do a good job of asking about your needs? Why or why not?”

GPT-4 with Paper Plain, when compared to other models, was much longer in length and not simplifying enough. Therefore, I’m also testing a modified prompt that includes “Summarize” and another prompt that includes “Make it Short” Distribution of all the model generation lengths: