Research map / Architectures and efficiency / Pruning and efficient tuning
FlexLoRA: Entropy-Guided Flexible Low-Rank Adaptation
Muqing Liu, Chongjie Si, Yuheng Jia
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 |
|---|---|---|---|
| use entropy importance instead of nuclear norm (ablation) | FlexLoRA · GLUE dev | Avg. score: 87.7 -> 89.1 | Improved |
| use entropy importance instead of Frobenius norm (ablation) | FlexLoRA · GLUE dev | Avg. score: 87.1 -> 89.1 | Improved |
| add rank expansion to prune-only (ablation) | FlexLoRA · GLUE dev | Avg. score: 87.5 -> 89.1 | Improved |
| add rank pruning to expand-only (ablation) | FlexLoRA · GLUE dev | Avg. score: 87.6 -> 89.1 | Improved |
| use zero-impact init instead of small-init (ablation) | FlexLoRA · GLUE dev | Avg. score: 87.8 -> 89.1 | Improved |
| use zero-impact init instead of zero-init (ablation) | FlexLoRA · GLUE dev | Avg. score: 87.1 -> 89.1 | Improved |
| use zero-impact init instead of orthogonal-init (ablation) | FlexLoRA · GLUE dev | Avg. score: 88.0 -> 89.1 | Improved |
| FlexLoRA instead of AdaLoRA | DeBERTaV3-base · GLUE dev | Avg. score: 88.1 -> 89.1 | Improved |
| FlexLoRA instead of AdaLoRA at rank 8 | LLaMA3-8B · Commonsense reasoning | Avg. accuracy: 85.1 -> 85.2 | No clear effect |
| raise rank from 8 to 32 for FlexLoRA (ablation) | LLaMA3-8B · Commonsense reasoning | Avg. accuracy: 85.2 -> 85.5 | Improved |
| FlexLoRA instead of LoRA r=32 | LLaMA3-8B · Commonsense reasoning | Avg. accuracy: 85.4 -> 85.5 | No clear effect |
| FlexLoRA instead of LoRA r=14 | ViT-B/16 · VTAB-1K | Avg. accuracy: 66.7 -> 67.8 | Improved |
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.
- DoRA — DoRA: Weight-Decomposed Low-Rank Adaptation (ICML 2024) · better on 1 of 1 shared result
- PiSSA — PiSSA: Principal Singular Values and Singular Vectors Adaptation of Large Language Models (NeurIPS 2024) · better on 1 of 1 shared result
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
- SDS-LoRA: Overcoming Anisotropic Gradient Scaling in Low-Rank Adaptation (NeurIPS 2026 · similarity 0.91)
- Flat-LoRA: Low-Rank Adaptation over a Flat Loss Landscape (ICML 2025 · similarity 0.90)
- The Primacy of Magnitude in Low-Rank Adaptation (NeurIPS 2025 · similarity 0.90)
- GraLoRA: Granular Low-Rank Adaptation for Parameter-Efficient Fine-Tuning (NeurIPS 2025 · similarity 0.90)
- RandLoRA: Full rank parameter-efficient fine-tuning of large models (ICLR 2025 · similarity 0.90)
- RepLoRA: Reparameterizing Low-rank Adaptation via the Perspective of Mixture of Experts (ICML 2025 · similarity 0.89)
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.
