SOLAR: Communication-Efficient Model Adaptation via Subspace-Oriented Reparametrization

ICLR 2026 Conference Submission13820 Authors

Keywords: Parameter-efficient fine-tuning, Adapter compression, LoRA, Sparse approximation, Model adaptation, Communication-efficient training, Foundation models, Post-training compression

TL;DR: We propose SOLAR, a post-training compression framework that reparameterizes PEFT adapters as sparse combinations of model-informed subspace bases, reducing communication and storage costs without sacrificing performance.

Abstract:

Parameter-efficient fine-tuning (PEFT) methods, such as LoRA, enable scalable adaptation of foundation models by injecting low-rank adapters. However, their communication and storage costs remain a major bottleneck in resource-constrained settings.
We propose SOLAR (Subspace-Oriented Latent Adapter Reparameterization), a post-training compression framework that substantially reduces the communication cost (i.e., the number of parameters to transmit or store) of PEFT adapters.
SOLAR expresses each PEFT update as a linear combination of basis vectors formed from the foundation model’s singular vectors with controlled random perturbations.
By exploiting the subspace similarity (the alignment of principal directions) between the foundation model and task-specific fine-tuned updates, SOLAR decouples the adapter size from PEFT structure and ensures compact yet expressive representations. It is model-agnostic and compatible with existing PEFT methods, including LoRA and other adapter modules.
We theoretically establish a bound on the reconstruction error.
Experiments on language and vision tasks using LLaMA, GPT, and ViT models demonstrate that SOLAR preserves task performance while significantly reducing model representation sizes, offering an effective and communication-efficient solution for deployment in distributed systems and edge devices.

Primary Area: other topics in machine learning (i.e., none of the above)

Submission Number: 13820