A collection of recent papers, benchmarks and datasets of AI4Code domain.

machine-learning program-analysis code-representation

• Updated Apr 23, 2024

A powerful relational representation of source code

deep-learning pytorch relational-database semantic-analysis codeql code-representation

• Updated Sep 5, 2023
• Python

A Survey of Deep Learning Models for Structural Code Understanding

nlp deep-neural-networks code control-flow-graph code-generation abstract-syntax-tree gnn code-representation

• Updated May 12, 2022
• Python

Neural Data-Flow Analysis: A tool for solving program-related tasks which involve data-flow analysis using deep neural networks

deep-learning pytorch control-flow-graph abstract-syntax-tree data-flow-analysis program-dependence-graph ml4code ndfa gnn code-representation hierarchic-model

• Updated Nov 7, 2023
• Python

LLM-Lang: a compact transfer representation for code — canon. Sources compile into canon (.ir) for agents and emit back to JS/TS + C#, round-trip stable. ~57% fewer tokens on fresh code; behavior identity proven against a 61,581-test vendored suite.

javascript rust typescript csharp compiler transpiler developer-tools canon agents tokenization ai-agents round-trip code-representation llm

• Updated Oct 27, 2026
• Rust

distills Python repositories into compact review bundles for LLMs and agents

static-analysis agents repo-stats repository-analysis code-representation ai-code-review llm-agents llm-review context-optimization openclaw

• Updated Oct 28, 2026
• Python