Deciding when to escalate a query from a small language model to a larger one requires a cheap signal that predicts, before the large model is called, whethe…
机构:Cisco
来源:arXiv 2610.07354 | AI4Papers 论文推荐平台