Research map / Architectures and efficiency / Pruning and efficient tuning
E²LoRA: Efficient and Effective Low-Rank Adaptation with Entropy-Guided Adaptive Sharing
Minglei Li, Peng Ye, Jingqi Ye, haonan he, Tao Chen
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 entropy-guided sharing and rank allocation to LoRA | LoRA · GSM8K | accuracy: 70.21 -> 70.51 (6.82M -> 3.66M params) | Improved |
| add E2LoRA to LoRA | LoRA · HumanEval | Pass@1: 42.68 -> 44.31 | Improved |
| add E2LoRA to ShareLoRA | ShareLoRA · HumanEval | Pass@1: 43.69 -> 44.71 | Improved |
| add E2LoRA to ShareLoRA | ShareLoRA · GSM8K | accuracy: 70.51 -> 70.26 | No clear effect |
| add E2LoRA to LoRA | LoRA · GLUE (RoBERTa-Base) | average score: 85.87 -> 85.57 (0.3M -> 0.16M params) | No clear effect |
| add E2LoRA to ShareLoRA | ShareLoRA · GLUE (RoBERTa-Base) | average score: 84.99 -> 85.40 | Improved |
| add E2LoRA to LoRA | LoRA · CLIP image classification (7 datasets) | average accuracy: 89.08 -> 89.81 (1.31M -> 0.66M params) | Improved |
| add E2LoRA to ShareLoRA | ShareLoRA · CLIP image classification (7 datasets) | average accuracy: 88.75 -> 89.73 (0.72M -> 0.39M params) | Improved |
| add heterogeneity-based rank allocation (HRA) on top of LSS (ablation) | E2LoRA · GSM8K (Llama-3.1-8B) | accuracy: 69.45 -> 70.51 | Improved |
| add local similarity-based sharing (LSS) on top of HRA (ablation) | E2LoRA · GSM8K (Llama-3.1-8B) | accuracy: 71.19 -> 70.51 (6.82M -> 3.61M params) | Mixed |
| use proxy entropy similarity instead of cosine similarity (ablation) | E2LoRA · GSM8K (Llama-3.1-8B) | accuracy: 66.67 -> 70.51 | Improved |
| use proxy entropy similarity instead of KL similarity (ablation) | E2LoRA · GSM8K (Llama-3.1-8B) | accuracy: 67.57 -> 70.51 | 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.
- VeRA — VeRA: Vector-based Random Matrix Adaptation (ICLR 2024) · better on 2 of 2 shared results
- DoRA — DoRA: Weight-Decomposed Low-Rank Adaptation (ICML 2024) · better on 2 of 2 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
- BSLoRA: Enhancing the Parameter Efficiency of LoRA with Intra-Layer and Inter-Layer Sharing (ICML 2025 · similarity 0.90)
- RepLoRA: Reparameterizing Low-rank Adaptation via the Perspective of Mixture of Experts (ICML 2025 · similarity 0.89)
- SDS-LoRA: Overcoming Anisotropic Gradient Scaling in Low-Rank Adaptation (NeurIPS 2026 · similarity 0.89)
- GraLoRA: Granular Low-Rank Adaptation for Parameter-Efficient Fine-Tuning (NeurIPS 2025 · similarity 0.89)
- Flat-LoRA: Low-Rank Adaptation over a Flat Loss Landscape (ICML 2025 · similarity 0.89)
- AuroRA: Breaking Low-Rank Bottleneck of LoRA with Nonlinear Mapping (NeurIPS 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.
