1
GPart: End-to-End Isometric Fine-Tuning via Global Parameter Partitioning
LoRA的等距瓶颈终于被打破！GPart用单次随机投影实现端到端等距微调，仅需存储d+1个值，在NLU、CV、数学推理上达到SOTA，参数效率极致。
arXiv:2605.14841v1 Announce Type: cross Abstract: Low-rank adaptation (LoRA) has become the dominant paradigm for parameter-efficient fine-tuning (PEF…