Research map / Large language models / Instruction and fine-tuning
BA-LoRA: Bias-Alleviating Low-Rank Adaptation to Mitigate Catastrophic Inheritance in Large Language Models
Yupeng Chang, Yi Chang, Yuan Wu
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 consistency regularizer (ablation) | PiSSA · GSM8K | accuracy: 51.48 -> 54.25 | Improved |
| add diversity regularizer (ablation) | PiSSA · GSM8K | accuracy: 51.48 -> 53.60 | Improved |
| add SVD regularizer (ablation) | PiSSA · GSM8K | accuracy: 51.48 -> 52.95 | Improved |
| add all three regularizers (ablation) | PiSSA · GSM8K | accuracy: 51.48 -> 55.86 | Improved |
| add consistency regularizer (ablation) | PiSSA · MATH | accuracy: 7.60 -> 9.15 | Improved |
| add all three regularizers (ablation) | PiSSA · GLUE average (DeBERTa-v3-base) | average score: 89.47 -> 90.67 | Improved |
| add three regularizers (ablation) | LoRA · GSM8K | accuracy: 42.68 -> 51.82 | Improved |
| add three regularizers (ablation) | DoRA · GSM8K | accuracy: 41.77 -> 52.71 | Improved |
| add three regularizers (ablation) | LoRA · MBPP | Pass@1: 21.51 -> 33.81 | Improved |
| use covariance diversity regularizer instead of entropy (ablation) | BA-LoRA · MNLI | accuracy: 90.41 -> 91.26 | Improved |
| BA-LoRA vs PiSSA on noisier-pretrained T5-base | BA-LoRA · GLUE subset | average score: 84.71 -> 87.97 | Improved |
| BA-LoRA vs PiSSA on RoBERTa-base | BA-LoRA · GLUE subset | average score: 85.23 -> 86.34 | Improved |
Closest papers by meaning
- Beyond LoRA vs. Full Fine-Tuning: Gradient-Guided Optimizer Routing for LLM Adaptation (NeurIPS 2026 · similarity 0.91)
- GoRA: Gradient-driven Adaptive Low Rank Adaptation (NeurIPS 2025 · similarity 0.91)
- C-LoRA: Contextual Low-Rank Adaptation for Uncertainty Estimation in Large Language Models (NeurIPS 2025 · similarity 0.91)
- Learning Rate Matters: Vanilla LoRA May Suffice for LLM Fine-tuning (NeurIPS 2026 · similarity 0.90)
- RefLoRA: Refactored Low-Rank Adaptation for Efficient Fine-Tuning of Large Models (NeurIPS 2025 · similarity 0.90)
- LoRA-One: One-Step Full Gradient Could Suffice for Fine-Tuning Large Language Models, Provably and Efficiently (ICML 2025 · similarity 0.90)
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.
