Research map / Large language models / Retrieval and knowledge
SRACG: A Code Generation Framework with Selective Retrieval Augmentation
Mengzhen Wang, Shukai Ma, Songwen Gong, Jiexin Wang, Ruolin Chen, Liuwen Cao, Yi Cai
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 |
|---|---|---|---|
| add necessity-aware intent selection (ablation) | SRACG (GPT-3.5-Turbo) · HumanEval+ | Pass@1: 62.37 -> 66.88 | Improved |
| add necessity-aware intent selection (ablation) | SRACG (DeepSeek-V3) · HumanEval+ | Pass@1: 82.11 -> 85.97 | Improved |
| add multi-objective retrieval (ablation) | SRACG (GPT-3.5-Turbo) · HumanEval+ | Pass@1: 63.71 -> 66.88 | Improved |
| add preference consistency filtering (ablation) | SRACG (GPT-3.5-Turbo) · HumanEval+ | Pass@1: 63.50 -> 66.88 | Improved |
| add preference consistency filtering (ablation) | SRACG (DeepSeek-V3) · HumanEval+ | Pass@1: 83.11 -> 85.97 | Improved |
| add execution plan extraction (ablation) | SRACG (GPT-3.5-Turbo) · HumanEval+ | Pass@1: 64.41 -> 66.88 | Improved |
| add execution plan extraction (ablation) | SRACG (DeepSeek-V3) · HumanEval+ | Pass@1: 83.91 -> 85.97 | Improved |
| indiscriminate top-1 retrieval augmentation (RACG) vs no retrieval | GPT-3.5-Turbo · HumanEval+ | Pass@1: 59.82 -> 56.21 | Hurt |
| indiscriminate top-1 retrieval augmentation (RACG) vs no retrieval | DeepSeek-V3 · HumanEval+ | Pass@1: 82.80 -> 80.22 | Hurt |
| apply SRACG to base LLM | GPT-3.5-Turbo · HumanEval+ | Pass@1: 59.82 -> 66.88 | Improved |
| apply SRACG to base LLM | DeepSeek-V3 · HumanEval+ | Pass@1: 82.80 -> 85.97 | Improved |
| apply SRACG to base LLM | GPT-3.5-Turbo · CodeContests | Pass@1: 2.39 -> 5.61 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- CODET — CodeT: Code Generation with Generated Tests (ICLR 2023) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Repoformer: Selective Retrieval for Repository-Level Code Completion (ICML 2024 · similarity 0.91)
- Let's Revise Step-by-Step: A Unified Local Search Framework for Code Generation with LLMs (NeurIPS 2025 · similarity 0.89)
- PinNet: Pinpoint Instructive Information for Retrieval Augmented Code-to-Text Generation (ICML 2024 · similarity 0.88)
- CodeMMR: Bridging Natural Language, Code, and Image for Unified Retrieval (CVPR 2026 · similarity 0.88)
- EffiCoder: Enhancing Code Generation in Large Language Models through Efficiency-Aware Fine-tuning (ICML 2025 · similarity 0.87)
- Planning in Natural Language Improves LLM Search for Code Generation (ICLR 2025 · similarity 0.87)
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.
