Improve Large Language Model Systems with User Logs

Changyue Wang, Weihang Su, Qingyao Ai +4

Scaling training data and model parameters has long driven progress in large language models (LLMs), but this paradigm is increasingly constrained by the scarcity of high-quality d…