Hi, it was a great project with valuable dataset and codebase. I have questions when reading the paper and code, how many RL training have you done? I am asking this question because I wonder how much cost it will take to train the model with such high-quality data (large in size, multi-turn search). To my knowledge, the search API is quiet expensive, did you find some way to reduce the cost?
Hi, it was a great project with valuable dataset and codebase. I have questions when reading the paper and code, how many RL training have you done? I am asking this question because I wonder how much cost it will take to train the model with such high-quality data (large in size, multi-turn search). To my knowledge, the search API is quiet expensive, did you find some way to reduce the cost?