Research map / Architectures and efficiency / Pruning and efficient tuning
IsoBN: Fine-Tuning BERT with Isotropic Batch Normalization
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Apply IsoBN before the classifier | BERT-base fine-tuning · GLUE dev average | Avg: 81.37 -> 82.36 | Helped |
| Apply IsoBN before the classifier | RoBERTa-large fine-tuning · GLUE dev average | Avg: 88.16 -> 88.98 | Helped |
| Use correlation-aware IsoBN instead of batch normalization (ablation) | BERT-base embeddings · MRPC | EV_3: 0.86 -> 0.77 | Helped |
| Use correlation-aware IsoBN instead of batch normalization (ablation) | RoBERTa-large embeddings · MRPC | EV_3: 0.76 -> 0.43 | Helped |
Closest papers by meaning
- Stable Anisotropic Regularization (ICLR 2024 · similarity 0.89)
- Revisiting Few-sample BERT Fine-tuning (ICLR 2021 · similarity 0.87)
- Isotropy in the Contextual Embedding Space: Clusters and Manifolds (ICLR 2021 · similarity 0.87)
- Mixout: Effective Regularization to Finetune Large-scale Pretrained Language Models (ICLR 2020 · similarity 0.87)
- On the Stability of Fine-tuning BERT: Misconceptions, Explanations, and Strong Baselines (ICLR 2021 · similarity 0.87)
- SimReg: Achieving Higher Performance in the Pretraining via Embedding Similarity Regularization (NeurIPS 2026 · similarity 0.86)
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.