Comment by yifanl

Is it just an open secret that the models are currently just being hardcoded for random benchmarks? Seems weird that people would be asking Putnam problems to a chatbot :/