Research map / Architectures and efficiency / Attention and transformers
GraphCodeBERT: Pre-training Code Representations with Data Flow
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 |
|---|---|---|---|
| Add data flow to code representation (ablation) | GraphCodeBERT · CodeSearchNet code search (Overall) | MRR: 0.713 -> 0.693 | Helped |
| Add Edge Prediction pre-training (ablation) | GraphCodeBERT · CodeSearchNet code search (Overall) | MRR: 0.707 -> 0.713 | Helped |
| Add Node Alignment pre-training (ablation) | GraphCodeBERT · CodeSearchNet code search (Overall) | MRR: 0.703 -> 0.713 | Helped |
| Use data-flow structure and structure-aware pre-training | GraphCodeBERT versus CodeBERT · CodeSearchNet code search (Overall) | MRR: 0.693 -> 0.713 | Helped |
| Use data-flow structure | GraphCodeBERT versus CodeBERT · Code refinement (medium) | Acc: 5.2 -> 9.1 | Helped |
Closest papers by meaning
- Global Relational Models of Source Code (ICLR 2020 · similarity 0.91)
- AST-T5: Structure-Aware Pretraining for Code Generation and Understanding (ICML 2024 · similarity 0.88)
- CODE REPRESENTATION LEARNING AT SCALE (ICLR 2024 · similarity 0.88)
- Towards A Generalist Code Embedding Model Based On Massive Data Synthesis (NeurIPS 2025 · similarity 0.86)
- Exploiting Code Symmetries for Learning Program Semantics (ICML 2024 · similarity 0.86)
- ObscuraCoder: Powering Efficient Code LM Pre-Training Via Obfuscation Grounding (ICLR 2025 · 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.