Research map / Image generation / Image restoration
Code Generation as a Dual Task of Code Summarization
Bolin Wei, Ge Li, Xin Xia, Zhiyi Fu, Zhi Jin
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 |
|---|---|---|---|
| jointly train CS and CG with both duality constraints | Seq2Seq CS and CG models · Java | BLEU: 41.01 -> 42.39 | Improved |
| jointly train CS and CG with both duality constraints | Seq2Seq CS and CG models · Python | BLEU: 20.47 -> 21.80 | Improved |
| jointly train CS and CG with both duality constraints | Seq2Seq CS and CG models · Java | CG BLEU: 10.86 -> 17.17 | Improved |
| jointly train CS and CG with both duality constraints | Seq2Seq CS and CG models · Python | CG BLEU: 10.43 -> 12.09 | Improved |
| add probabilistic duality regularization only (ablation) | CS model · Java | BLEU: 41.01 -> 41.73 | Improved |
| add probabilistic duality regularization only (ablation) | CS model · Python | BLEU: 20.47 -> 21.66 | Improved |
| add attention duality regularization only (ablation) | CS model · Java | BLEU: 41.01 -> 41.96 | Improved |
| add attention duality regularization only (ablation) | CS model · Python | BLEU: 20.47 -> 21.57 | Improved |
| combine probabilistic and attention duality regularization rather than using either alone (ablation) | CS model · Java | BLEU: 41.73 -> 42.39; 41.96 -> 42.39 | Improved |
| combine probabilistic and attention duality regularization rather than using either alone (ablation) | CS model · Python | BLEU: 21.66 -> 21.80; 21.57 -> 21.80 | Improved |
| use attention duality rather than probabilistic duality alone (ablation) | CS model · Java | BLEU: 41.73 -> 41.96 | Improved |
| use attention duality rather than probabilistic duality alone (ablation) | CS model · Python | BLEU: 21.66 -> 21.57 | No clear effect |
Closest papers by meaning
- Retrieval-Augmented Generation for Code Summarization via Hybrid GNN (ICLR 2021 · similarity 0.87)
- PinNet: Pinpoint Instructive Information for Retrieval Augmented Code-to-Text Generation (ICML 2024 · similarity 0.82)
- Language Models of Code Are Few-Shot Planners and Reasoners for Multi-Document Summarization with Attribution (AAAI 2025 · similarity 0.82)
- JanusCoder: Towards a Foundational Visual-Programmatic Interface for Code Intelligence (ICLR 2026 · similarity 0.82)
- CODE REPRESENTATION LEARNING AT SCALE (ICLR 2024 · similarity 0.82)
- Primal Generation, Dual Judgment: Self-Training from Test-Time Scaling (NeurIPS 2026 · similarity 0.81)
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.
