Research map / Architectures and efficiency / Pruning and efficient tuning
CSPLoRA: Confidence-Guided Structure Planning for Low-Rank Adaptation
Huiming Ding, Xiaochen Li, Jianhui Ma, Xu An, Yihui Yang, Zhenyu Tan
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 |
|---|---|---|---|
| plan ranks with CSPLoRA instead of uniform rank | LoRA on LLaMA-2-7B · Commonsense reasoning avg | accuracy: 73.49 -> 74.74 | Improved |
| plan ranks with CSPLoRA instead of uniform rank | DoRA on LLaMA-2-7B · Commonsense reasoning avg | accuracy: 73.48 -> 74.91 | Improved |
| plan ranks with CSPLoRA instead of uniform rank | PiSSA on LLaMA-2-7B · Commonsense reasoning avg | accuracy: 76.25 -> 76.97 | Improved |
| plan ranks with CSPLoRA instead of uniform rank | LoRA+ on LLaMA-3.1-8B · Commonsense reasoning avg | accuracy: 84.69 -> 84.44 | Hurt |
| plan ranks with CSPLoRA instead of uniform rank | LoRA on RoBERTa-large · GLUE dev avg | score: 85.30 -> 86.21 (+0.91) | Improved |
| plan ranks with CSPLoRA instead of uniform rank | DoRA on RoBERTa-large · GLUE dev avg | score: 88.62 -> 88.65 | No clear effect |
| plan ranks with CSPLoRA instead of uniform rank | PiSSA on LLaMA-2-7B · GSM8K | accuracy: 41.42 -> 43.77 | Improved |
| plan ranks with CSPLoRA instead of uniform rank | LoRA+ on LLaMA-3.1-8B · MATH | accuracy: 24.33 -> 24.26 | Hurt |
| add logarithmic smoothing of saliency scores (ablation) | CSPLoRA · Commonsense 5-task subset | average accuracy: 73.08 -> 73.51 | Improved |
| add score calibration (ablation) | CSPLoRA · Commonsense 5-task subset | average accuracy: 73.18 -> 73.51 | Improved |
| use sum Taylor saliency instead of Fisher information (ablation) | CSPLoRA · Commonsense 5-task subset | average accuracy: 73.37 -> 73.51 | Improved |
| use sum Taylor instead of mean Taylor saliency (ablation) | CSPLoRA · Commonsense 5-task subset | average accuracy: 73.47 -> 73.51 | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- PiSSA — PiSSA: Principal Singular Values and Singular Vectors Adaptation of Large Language Models (NeurIPS 2024) · better on 4 of 4 shared results
- DoRA — DoRA: Weight-Decomposed Low-Rank Adaptation (ICML 2024) · better on 3 of 3 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
- A Stronger Mixture of Low-Rank Experts for Fine-Tuning Foundation Models (ICML 2025 · similarity 0.88)
- LoRA Training Provably Converges to a Low-Rank Global Minimum Or It Fails Loudly (But it Probably Won't Fail) (ICML 2025 · similarity 0.88)
- LoRA-Pro: Are Low-Rank Adapters Properly Optimized? (ICLR 2025 · similarity 0.88)
- Come Together, But Not Right Now: A Progressive Strategy to Boost Low-Rank Adaptation (ICML 2025 · similarity 0.88)
- SeedLoRA: A Fusion Approach to Efficient LLM Fine-Tuning (ICML 2025 · similarity 0.88)
- AuroRA: Breaking Low-Rank Bottleneck of LoRA with Nonlinear Mapping (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.
