Structured pruning reduces the deployment cost of reasoning language models, but the resulting capability degradation can hinder subsequent on-policy distill…
机构:腾讯
来源:arXiv 2610.11332 | AI4Papers 论文推荐平台