CoCoMUT: A Tool for Code-Context Mining and Automated Dataset Generation

2026-06-26 · Source: cs.SE updates on arXiv.org · Field: Technology & Digital — Software Development & Engineering, Artificial Intelligence & Machine Learning, Data Science & Analytics · Depth: Expert, long

Summary

CoCoMUT is a Java tool designed for automated code-context mining and dataset generation, addressing the challenge of manually collecting method-level context for software-engineering assistants. It extracts comprehensive context for focal methods or generates datasets at class, package, or system scope. The tool discovers project structure, resolves build and classpath information, constructs a SootUp static call graph, and reconciles bytecode-level call edges with Spoon-based source extraction. Each method record combines source, class, documentation, call-graph, and metadata context, providing reproducible inputs for training and running learned software-engineering techniques. Evaluated on 20 real-world Java repositories (10 Maven, 10 Gradle), CoCoMUT processed all, emitting 56,512 method-context records and 386,048 serialized call edges. It achieved a 97.8% reconciliation rate for project-source bytecode targets and a 99.0% pass rate in a 200-record manual audit.

Key takeaway

For AI Scientists and Software Engineers developing context-aware software engineering techniques, CoCoMUT offers a robust, reproducible foundation for generating method-level context datasets. You can leverage its automated pipeline to create high-quality, versioned JSONL records, significantly reducing manual data collection efforts and improving the reliability of your model training and evaluation. Consider integrating CoCoMUT to standardize context extraction across diverse Java projects.

Key insights

CoCoMUT unifies source and bytecode analysis to generate reproducible, comprehensive method-level context datasets for Java projects.

Principles

• Deterministic source identity prevents false links.
• Unify source and bytecode views for complete context.
• Versioned JSONL outputs ensure reproducibility.

Method

CoCoMUT's pipeline involves project discovery, build/classpath resolution, Spoon source analysis, SootUp call-graph construction, source–bytecode reconciliation, and versioned JSONL generation.

In practice

• Generate datasets for LLM-based task automation.
• Extract context for code explanation or repair.
• Audit generated records for quality assurance.

Topics

• Code Context Mining
• Java Program Analysis
• Dataset Generation
• Static Call Graph
• Source-Bytecode Reconciliation
• Software Engineering Tools

Code references

• assert-lab/CoCoMUT
• fasterxml/jackson

Best for: AI Scientist, Research Scientist, Software Engineer

Related on AIssential

• iCoRe: An Iterative Correlation-Aware Retriever for Bug Reproduction Test Generation — iCoRe enhances bug reproduction test generation via correlation-aware, iterative context retrieval, differentiating code/tests and integrating function call analysis.
• Context Graphs and Event-Driven Architectures — The article introduces context graphs as graph-oriented, append-only logs of events, leveraging RDF-Star reification to capture comprehensive metadata for each assertion.
• Code Concepts: A Large-Scale Synthetic Dataset Generated from Programming Concept Seeds — Concept-driven synthetic data generation significantly improves LLM programming proficiency on benchmarks like HumanEval.
• cocoindex-io / cocoindex — CocoIndex provides incremental data pipelines for AI agents, ensuring fresh context with minimal reprocessing.
• Microskill Architecture: A Modular Skill-Driven Framework for AI-Native Code Generation — MicroSkill Architecture uses modular, semantically-routed skill capsules to optimize context and improve AI-native code generation efficiency.
• Project-wise Comparison of Software Birthmarks Using Weighted Partial Similarity — Project-wise software birthmark comparison requires weighting module similarities and focusing on top matches to detect partial reuse and mitigate incidental similarity.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.SE updates on arXiv.org.