English

HD-PiSSA: High-Rank Distributed Orthogonal Adaptation

Machine Learning 2025-09-29 v3 Artificial Intelligence

Abstract

Existing parameter-efficient fine-tuning (PEFT) methods for large language models (LLMs), such as LoRA and PiSSA, constrain model updates to low-rank subspaces, limiting their expressiveness and leading to suboptimal performance on complex tasks. To address this, we introduce High-rank Distributed PiSSA (HD-PiSSA), a distributed PEFT approach that initializes orthogonal adapters across different devices and aggregates their delta updates collectively on W for fine-tuning. Unlike Data Parallel LoRA or PiSSA, which maintain identical adapters across all devices, HD-PiSSA assigns different principal components of the pre-trained weights to each GPU, significantly expanding the range of update directions. This results in over 16x higher effective updated ranks than data-parallel LoRA or PiSSA when fine-tuning on 8 GPUs with the same per-device adapter rank. Empirically, we evaluate HD-PiSSA across various challenging downstream tasks, including mathematics, code generation, and multi-task learning. In the multi-task setting, HD-PiSSA achieves average gains of 10.0 absolute points (14.63%) over LoRA and 4.98 points (6.60%) over PiSSA across 12 benchmarks, demonstrating its benefits from the extra optimization flexibility.

Keywords

Cite

@article{arxiv.2505.18777,
  title  = {HD-PiSSA: High-Rank Distributed Orthogonal Adaptation},
  author = {Yiding Wang and Fauxu Meng and Xuefeng Zhang and Fan Jiang and Pingzhi Tang and Muhan Zhang},
  journal= {arXiv preprint arXiv:2505.18777},
  year   = {2025}
}
R2 v1 2026-07-01T02:36:10.506Z