1% VS 100%: Parameter-Efficient Low Rank Adapter for Dense Predictions | Read Paper on Bytez