Robert Lukoszko@Karmedge

2 years ago

Today local models took a hit https://t.co/uPQYkw8fQk

Kyle Corbitt@corbtt

2 years ago

So gpt-4o mini was absolutely dominating gpt-4o on our internal llm-as-judge evals. So I LOOKED AT THE DATA (h/t @HamelHusain) and realized it was answering the question BUT ALSO throwing in a bunch of random extra information that the prompt in no way asked for, and the judge…

Greg Speicher@Greg_Speicher

2 years ago

Robotics won’t have a ChatGPT Moment https://t.co/wnImo55r8F