Research map / Architectures and efficiency / Pruning and efficient tuning
OPLoRA: Orthogonal Projection LoRA Prevents Catastrophic Forgetting During Parameter-Efficient Fine-Tuning
Yifeng Xiong, Xiaohui Xie
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 |
|---|---|---|---|
| replace LoRA with OPLoRA-128 | LLaMA-2 7B · commonsense avg (fine-tune + forgetting) | average accuracy: 57.03 -> 57.76 | Improved |
| replace LoRA with OPLoRA-16 | Qwen2.5 7B · commonsense avg (fine-tune + forgetting) | average accuracy: 63.24 -> 64.26 | Improved |
| replace LoRA with OPLoRA-128 (forgetting MBPP) | LLaMA-2 7B · MBPP after commonsense fine-tuning | accuracy: 20.2 -> 20.4 | Improved |
| replace LoRA-Null with OPLoRA-128 (forgetting MBPP) | LLaMA-2 7B · MBPP after commonsense fine-tuning | accuracy: 9.8 -> 20.4 | Improved |
| replace LoRA with OPLoRA-16 | LLaMA-2 7B · GSM8K | accuracy: 44.88 -> 49.73 | Improved |
| replace LoRA with OPLoRA-16 | Qwen2.5 7B · GSM8K | accuracy: 82.18 -> 83.70 | Improved |
| replace LoRA with OPLoRA-16 | LLaMA-2 7B · MBPP | pass rate (%): 37.6 -> 40.5 | Improved |
| replace LoRA with OPLoRA-16 | LLaMA-2 7B · MBPP++ | pass rate (%): 30.4 -> 34.1 | Improved |
| replace LoRA with OPLoRA-16 | Qwen2.5 7B · MBPP | pass rate (%): 80.2 -> 80.2 | No clear effect |
| use PiSSA instead of LoRA | LLaMA-2 7B · commonsense avg (fine-tune + forgetting) | average accuracy: 57.03 -> 32.42 | Hurt |
| use projection rank k=128 instead of k=16 | OPLoRA · tasks overall | accuracy | Mixed |
| use OPLoRA instead of LoRA (training cost) | LLaMA-2 7B commonsense · training time | hours: 5 h 13 m -> 6 h 12 m | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- PiSSA — PiSSA: Principal Singular Values and Singular Vectors Adaptation of Large Language Models (NeurIPS 2024) · better on 6 of 6 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- LoRA vs Full Fine-tuning: An Illusion of Equivalence (NeurIPS 2025 · similarity 0.90)
- Beyond Higher Rank: Token-wise Input-Output Projections for Efficient Low-Rank Adaptation (NeurIPS 2025 · similarity 0.90)
- RefLoRA: Refactored Low-Rank Adaptation for Efficient Fine-Tuning of Large Models (NeurIPS 2025 · similarity 0.89)
- AltLoRA: Towards Better Gradient Approximation in Low-Rank Adaptation with Alternating Projections (NeurIPS 2025 · similarity 0.89)
- LoRASculpt: Sculpting LoRA for Harmonizing General and Specialized Knowledge in Multimodal Large Language Models (CVPR 2025 · similarity 0.88)
- Uni-LoRA: One Vector is All You Need (NeurIPS 2025 · similarity 0.88)
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.
