Research map / Label-efficient and robust learning / Self-supervised and contrastive
Self-Supervised Bug Detection and Repair
Miltiadis Allamanis, Henry Richard Jackson-Flux, Marc Brockschmidt
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 |
|---|---|---|---|
| Replace random bug insertion with a learned adversarial bug selector (ablation) | PyBugLab · PyPIBugs | Joint accuracy: 21.2 -> 24.2 | Improved |
| Replace random bug insertion with a learned adversarial bug selector (ablation) | PyBugLab · PyPIBugs | Joint accuracy: 20.6 -> 24.0 | Improved |
| Add semantics-preserving code augmentation (ablation) | PyBugLab GNN · PyPIBugs | Joint accuracy: 24.2 -> 26.4 | Improved |
| Add semantics-preserving code augmentation (ablation) | PyBugLab GREAT · PyPIBugs | Joint accuracy: 24.0 -> 23.2 | Hurt |
| Add semantics-preserving code augmentation (ablation) | PyBugLab GNN · RandomBugs | Joint accuracy: 69.6 -> 70.3 | Improved |
| Add semantics-preserving code augmentation (ablation) | PyBugLab GREAT · RandomBugs | Joint accuracy: 64.0 -> 65.3 | Improved |
| Use a learned bug selector rather than random bug insertion (ablation) | PyBugLab · RandomBugs | Joint accuracy: 69.4 -> 69.6 | No clear effect |
| Use a learned bug selector rather than random bug insertion (ablation) | PyBugLab · RandomBugs | Joint accuracy: 63.9 -> 64.0 | No clear effect |
| Use GNN instead of GREAT for code representations | PyBugLab+Aug · PyPIBugs | Joint accuracy: 23.2 -> 26.4 | Mixed |
| Use PyBugLab rather than CuBERT | Bug detection · PyPIBugs-WrongOp | F1: 0.378 -> 0.746 | Improved |
| Use PyBugLab rather than CuBERT | Bug detection · PyPIBugs-VarMisuse | F1: 0.493 -> 0.787 | Improved |
Closest papers by meaning
- Less Is More: Adaptive Program Repair with Bug Localization and Preference Learning (AAAI 2025 · similarity 0.89)
- Counterexample Guided Program Repair Using Zero-Shot Learning and MaxSAT-based Fault Localization (AAAI 2025 · similarity 0.86)
- Teaching Large Language Models to Self-Debug (ICLR 2024 · similarity 0.84)
- CoRNStack: High-Quality Contrastive Data for Better Code Retrieval and Reranking (ICLR 2025 · similarity 0.84)
- LeDex: Training LLMs to Better Self-Debug and Explain Code (NeurIPS 2024 · similarity 0.84)
- RepoAudit: An Autonomous LLM-Agent for Repository-Level Code Auditing (ICML 2025 · similarity 0.83)
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.
