Full fine-tuning has achieved remarkable success on reasoning-intensive tasks, but it often requires substantial GPU resource consumption. Parameter-efficien…
机构:中科院
来源:arXiv 2610.29134 | AI4Papers 论文推荐平台