Research map / Large language models / Instruction and fine-tuning
ScaLoRA: Optimally Scaled Low-Rank Adaptation for Efficient High-Rank Fine-Tuning
Yilang Zhang, Xiaodong Yang, Yiwei Cai, Georgios B. Giannakis
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add column-wise scaling on top of scalar scaling (ablation) | ScaLoRA-I · GSM8K (Gemma-3-12B) | accuracy: 81.27 -> 82.11 | Improved |
| add column-wise scaling on top of scalar scaling (ablation) | ScaLoRA-I · MATH (Gemma-3-12B) | accuracy: 37.90 -> 37.96 | Improved |
| replace LoRA with ScaLoRA-I | LoRA · GSM8K (Gemma-3-12B) | accuracy: 81.20 -> 82.11 | Improved |
| replace LoRA with ScaLoRA | LoRA · GLUE average (DeBERTaV3-base) | average score: 88.13 -> 88.98 | Improved |
| replace LoRA with ScaLoRA | LoRA · GLUE RTE | accuracy: 84.48 -> 87.61 | Improved |
| replace LoRA with ScaLoRA at r=8 | LoRA · Commonsense average (LLaMA2-7B) | accuracy: 73.63 -> 74.51 | Improved |
| replace LoRA with ScaLoRA at r=8 | LoRA · Commonsense average (LLaMA3-8B) | accuracy: 76.83 -> 77.85 | Improved |
| use intermittent ScaLoRA-I (I=10) instead of ScaLoRA | ScaLoRA · Commonsense average (LLaMA2-7B) | accuracy: 74.51 -> 74.75 | No clear effect |
| use intermittent ScaLoRA-I (I=10) instead of ScaLoRA | ScaLoRA · Commonsense average (LLaMA3-8B) | accuracy: 77.85 -> 77.57 | Hurt |
| increase LoRA rank from 8 to 32 | LoRA · Commonsense average (LLaMA2-7B) | accuracy: 73.63 -> 74.52 | Improved |
| ScaLoRA at r=8 vs LoRA rank 32 | ScaLoRA · Commonsense average (LLaMA3-8B) | accuracy: 77.54 -> 77.85 | Improved |
| use ScaLoRA instead of LoRA for training cost | LoRA · Extra memory (LLaMA3-8B) | GPU memory: 0.01 GB extra vs 1.08 and 7.83 GB for MoRA and HiRA | Mixed |
Closest papers by meaning
- GoRA: Gradient-driven Adaptive Low Rank Adaptation (NeurIPS 2025 · similarity 0.94)
- RefLoRA: Refactored Low-Rank Adaptation for Efficient Fine-Tuning of Large Models (NeurIPS 2025 · similarity 0.94)
- Beyond Higher Rank: Token-wise Input-Output Projections for Efficient Low-Rank Adaptation (NeurIPS 2025 · similarity 0.93)
- Uni-LoRA: One Vector is All You Need (NeurIPS 2025 · similarity 0.93)
- LoRA-One: One-Step Full Gradient Could Suffice for Fine-Tuning Large Language Models, Provably and Efficiently (ICML 2025 · similarity 0.93)
- Data Efficient Adaptation in Large Language Models via Continuous Low-Rank Fine-Tuning (NeurIPS 2025 · similarity 0.93)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
