cs.LGMay 29, 2026

Finer Parameter Steps for Low-Rank PEFT: A Controlled Study with CP Tensor Adapters

Authors: Xinjue WangXiuheng WangYejun ZhangSergiy A. VorobyovEsa OllilaZhi-Yong Wang

Abstract

Low-rank adapters are usually compared by sweeping a small set of ranks, but the rank also fixes the resolution of the parameter budget. For a 2048×20482048{\times}2048 OPT attention projection, increasing LoRA by one rank stores 40964096 trainable scalars, leaving large gaps between feasible low-budget adapter sizes. This paper asks whether a tensorized adapter with finer capacity increments changes the observed accuracy--budget trade-off. We instantiate this question with fixed-component canonical polyadic (CP) tensor adapters. Under a 32×64×32×6432{\times}64{\times}32{\times}64 tensorization, one normalized CP component stores 193193 trainable scalars per projection, about 2121 times smaller than one LoRA rank step. We compare CP adapters and LoRA on OPT-1.3B across SST-2, RTE, and BoolQ under matched target modules, training protocol, data caps, and seed schedules. CP trains stably and fills the gaps between LoRA ranks, but the effect is task-dependent: SST-2 reaches an early low-budget plateau, BoolQ benefits from additional CP components before saturating slightly below LoRA, and RTE remains LoRA-favored. Finer parameter steps are therefore useful for diagnosing PEFT budget sensitivity, but they do not by themselves guarantee a better accuracy--budget curve.

Explore similar work

CardsList