3 ms·CausalLM is not optimal for in-context learning2 points by Jimmc414 3y agocubefox 3y agoDo modern language models already use prefix rather causal in-context learning?