Research map / Large language models / Language, safety and interpretability
StructBERT: Incorporating Language Structures into Pre-training for Deep Language Understanding
Wei Wang, Bin Bi, Ming Yan, Chen Wu, Jiangnan Xia, Zuyi Bao, Liwei Peng, Luo Si
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 word- and sentence-structural pre-training objectives to RoBERTa | RoBERTa · GLUE test | Average: 88.8 -> 89.0 | Improved |
| Add word- and sentence-structural pre-training objectives | BERT Large · SNLI test | Accuracy: 91.3 -> 91.7 | Improved |
| Add word- and sentence-structural pre-training objectives (ablation) | BERT Base · CoLA | Accuracy: 81.7 -> 85.8 | Improved |
| Add word-structural objective (ablation) | StructBERT Base · CoLA | Accuracy: 81.7 -> 85.8 | Improved |
| Add sentence-structural objective (ablation) | StructBERT Base · CoLA | Accuracy: 84.9 -> 85.8 | Improved |
| Add word-structural objective (ablation) | StructBERT Base · SST-2 | Accuracy: 92.7 -> 92.9 | Improved |
| Add sentence-structural objective (ablation) | StructBERT Base · SST-2 | Accuracy: 92.9 -> 92.9 | No clear effect |
| Add word-structural objective (ablation) | StructBERT Base · MNLI | Accuracy: 85.2 -> 85.4 | Improved |
| Add sentence-structural objective (ablation) | StructBERT Base · MNLI | Accuracy: 84.1 -> 85.4 | Improved |
| Add word-structural objective (ablation) | StructBERT Base · SNLI | Accuracy: 91.6 -> 91.5 | Hurt |
| Add sentence-structural objective (ablation) | StructBERT Base · SNLI | Accuracy: 91.1 -> 91.5 | Improved |
| Add word-structural objective (ablation) | StructBERT Base · QQP | Accuracy: 90.7 -> 91.1 | Improved |
Closest papers by meaning
- AutoBERT-Zero: Evolving BERT Backbone from Scratch (AAAI 2022 · similarity 0.87)
- XLNet: Generalized Autoregressive Pretraining for Language Understanding (NeurIPS 2019 · similarity 0.86)
- MosaicBERT: A Bidirectional Encoder Optimized for Fast Pretraining (NeurIPS 2023 · similarity 0.86)
- MPNet: Masked and Permuted Pre-training for Language Understanding (NeurIPS 2020 · similarity 0.86)
- DEBERTA: DECODING-ENHANCED BERT WITH DISENTANGLED ATTENTION (ICLR 2021 · similarity 0.86)
- Enhancing LLM’s Cognition via Structurization (NeurIPS 2024 · similarity 0.85)
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.
