AI RESEARCH
CeRA: Breaking the Linear Ceiling of Low-Rank Adaptation via Manifold Expansion
arXiv CS.LG
•
ArXi:2602.22911v3 Announce Type: replace Low-Rank Adaptation (LoRA) dominates parameter-efficient fine-tuning (PEFT). However, it faces a critical ``linear ceiling'' in complex reasoning tasks: simply increasing the rank yields diminishing returns due to intrinsic linear constraints. We