Intel / article

Better Call Grep: Evaluating and Improving Grep-Like Lexical Retrieval for Repository-Level Code Completion

Source: arxiv.org
Author: Baoyi Wang, Xingliang Wang, Guochang Li, Chen Zhi, Junxiao Han, Xinkui Zhao, Nan Wang, Shuiguang Deng, Jianwei Yin
Date: Published Jul 30, 2026

Context

An AI completing code often needs definitions from other files. GrepRAG finds that through text searches, avoiding a separately maintained search index. Its authors report that even a basic version, where the model generates search commands, competed with graph-based retrieval in their tests. The improved version ranks matches using code identifiers and removes structurally redundant results, addressing noisy matches and fragmented context. On CrossCodeEval, it achieved a reported 7.04–15.58% relative improvement in exact-match completion over the strongest comparison method. The finding makes lightweight search a credible baseline for repository code completion. Research abstract.

Key takeaways · AI-distilled

Letting the model write its own ripgrep commands to pull context matched graph-based retrieval that has to build and maintain a code index. It worked because the lines grep matched tended to sit physically near the spot being completed.
Plain text search breaks in two specific ways: high-frequency ambiguous keywords drag in noise, and fixed truncation boundaries cut retrieved code mid-structure so the model sees fragments.
GrepRAG fixes both with post-processing rather than a new index: re-rank matches by identifier weight, then dedupe along structural boundaries. That beat the best prior method by 7-15% relative on exact-match code completion.
The framing worth stealing: before paying for or a code graph, measure how far index-free search already gets you, since the index is a permanent maintenance cost and grep is not.

Terms in this piece · Glossary

context window — The maximum amount of text a model can consider at once — its working memory for the current conversation or task.
embedding — A list of numbers representing a piece of text's meaning, so that similar meanings end up numerically close and can be searched.

Key quotes

“Despite its simplicity, Naive GrepRAG achieves performance comparable to sophisticated graph-based baselines.”

“Further analysis shows that its effectiveness stems from retrieving lexically precise code fragments that are spatially closer to the completion site.”

“Extensive evaluation on CrossCodeEval and RepoEval-Updated demonstrates that GrepRAG consistently outperforms state-of-the-art (SOTA) methods, achieving 7.04-15.58 percent relative improvement in code exact match (EM) over the best baseline on CrossCodeEval.”

Read the source arxiv.org