# Researchers build source code knowledge graph spanning 145 million files

_Published Friday, September 25, 2026 at 11:07 AM EDT · AI, Science · Latest · Tier 2 — Notable_

Researchers have built CodeGraph, a knowledge graph that connects roughly 145 million source code files to concepts such as algorithms, design patterns and application domains. In a preprint, they report approximately 158 million nodes and 1 billion links across 14 programming languages.

The pipeline uses a code-focused language model to identify concepts, then links them to Wikidata so files can be related to broader categories. The researchers applied it to the 167 million files in the Stack-Edu corpus. Their quality checks combine a small human reference set with an AI judging filter.

## Sources

- [cs.CL updates on arXiv.org](https://arxiv.org/abs/2609.29474)

---
Canonical: https://techandbusiness.org/newswire/-X3nT_gddEtCBvhibycdsz
Published: 2026-09-25T15:07:35.421Z
Story chronology: 2026-09-25T04:00:00.000Z
Retrieved: 2026-09-25T17:06:09.145Z
Publisher: Tech & Business (techandbusiness.org)