Research map / Label-efficient and robust learning / Recognition and long-tail
Dynamic Neural Program Embeddings for Program Repair
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 |
|---|---|---|---|
| Use dynamic execution-trace embeddings instead of syntax-based token, AST, or run-time syntactic trace embeddings | Program error-pattern classification · Print Chessboard | accuracy: 26.3% -> 99.3% | Helped |
| Use dynamic execution-trace embeddings instead of syntax-based embeddings | Program error-pattern classification · Count Parentheses | accuracy: 25.5% -> 98.8% | Helped |
| Use dynamic execution-trace embeddings instead of syntax-based embeddings | Program error-pattern classification · Generate Binary Digits | accuracy: 23.8% -> 99.2% | Helped |
| Use prediction-guided dynamic embeddings rather than enumerative search | SarfGen · 3-5 fixes | repair time (seconds): 44.7 -> 3.1 | Helped |
| Use prediction-guided dynamic embeddings rather than enumerative search | SarfGen · 6-7 fixes | repair time (seconds): 95.9 -> 3.6 | Helped |
| Use prediction-guided dynamic embeddings rather than enumerative search | SarfGen · ≥8 fixes | repair time (seconds): 128.3 -> 38.8 | Helped |
| Use prediction-guided dynamic embeddings rather than enumerative search | SarfGen · 1-2 fixes | repair time (seconds): 3.8 -> 2.5; 3.8 -> 2.8; 3.8 -> 3.3 | Mixed |
| Use dependency enforcement rather than variable traces (ablation) | Dynamic program embedding · Print Chessboard | accuracy: 93.9% -> 99.3% | Helped |
| Use dependency enforcement rather than state traces (ablation) | Dynamic program embedding · Count Parentheses | accuracy: 93.8% -> 98.8% | Helped |
| Use dependency enforcement rather than state traces (ablation) | Dynamic program embedding · Generate Binary Digits | accuracy: 94.5% -> 99.2% | Helped |
Closest papers by meaning
- Neural Code Comprehension: A Learnable Representation of Code Semantics (NeurIPS 2018 · similarity 0.91)
- Counterexample Guided Program Repair Using Zero-Shot Learning and MaxSAT-based Fault Localization (AAAI 2025 · similarity 0.86)
- Emergent Representations of Program Semantics in Language Models Trained on Programs (ICML 2024 · similarity 0.86)
- CoRNStack: High-Quality Contrastive Data for Better Code Retrieval and Reranking (ICLR 2025 · similarity 0.86)
- Towards A Generalist Code Embedding Model Based On Massive Data Synthesis (NeurIPS 2025 · similarity 0.85)
- CODE REPRESENTATION LEARNING AT SCALE (ICLR 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.