Saved in:
| Main Authors: | , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2024
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2406.10263 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866914834930466816 |
|---|---|
| author | Zhang, Wenrui Fu, Tiehang Yuan, Ting Zhang, Ge Chen, Dong Wang, Jie |
| author_facet | Zhang, Wenrui Fu, Tiehang Yuan, Ting Zhang, Ge Chen, Dong Wang, Jie |
| contents | Recent advancements in Retrieval-Augmented Generation have significantly enhanced code completion at the repository level. Various RAG-based code completion systems are proposed based on different design choices. For instance, gaining more effectiveness at the cost of repeating the retrieval-generation process multiple times. However, the indiscriminate use of retrieval in current methods reveals issues in both efficiency and effectiveness, as a considerable portion of retrievals are unnecessary and may introduce unhelpful or even harmful suggestions to code language models. To address these challenges, we introduce CARD, a lightweight critique method designed to provide insights into the necessity of retrievals and select the optimal answer from multiple predictions. CARD can seamlessly integrate into any RAG-based code completion system. Our evaluation shows that CARD saves 21% to 46% times of retrieval for Line completion, 14% to 40% times of retrieval for API completion, and 6% to 46.5% times of retrieval for function completion respectively, while improving the accuracy. CARD reduces latency ranging from 16% to 83%. CARD is generalizable to different LMs, retrievers, and programming languages. It is lightweight with training in few seconds and inference in few milliseconds. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2406_10263 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | A Lightweight Framework for Adaptive Retrieval In Code Completion With Critique Model Zhang, Wenrui Fu, Tiehang Yuan, Ting Zhang, Ge Chen, Dong Wang, Jie Software Engineering Recent advancements in Retrieval-Augmented Generation have significantly enhanced code completion at the repository level. Various RAG-based code completion systems are proposed based on different design choices. For instance, gaining more effectiveness at the cost of repeating the retrieval-generation process multiple times. However, the indiscriminate use of retrieval in current methods reveals issues in both efficiency and effectiveness, as a considerable portion of retrievals are unnecessary and may introduce unhelpful or even harmful suggestions to code language models. To address these challenges, we introduce CARD, a lightweight critique method designed to provide insights into the necessity of retrievals and select the optimal answer from multiple predictions. CARD can seamlessly integrate into any RAG-based code completion system. Our evaluation shows that CARD saves 21% to 46% times of retrieval for Line completion, 14% to 40% times of retrieval for API completion, and 6% to 46.5% times of retrieval for function completion respectively, while improving the accuracy. CARD reduces latency ranging from 16% to 83%. CARD is generalizable to different LMs, retrievers, and programming languages. It is lightweight with training in few seconds and inference in few milliseconds. |
| title | A Lightweight Framework for Adaptive Retrieval In Code Completion With Critique Model |
| topic | Software Engineering |
| url | https://arxiv.org/abs/2406.10263 |