Retrieval-Augmented Generation for Scientific Code Understanding

• URL: http://arxiv.org/abs/2609.12190v1
• Date: Thu, 10 Sep 2026 20:28:38 GMT
• Title: Retrieval-Augmented Generation for Scientific Code Understanding
• Abstract summary: This work investigates whether a useful, fully local coding agent can be built around small open-source models by shifting the computational burden away from inference.<n>We develop a Retrieval-Augmented Generation (RAG) system for scientific code understanding that strictly separates an expensive offline ingestion stage parsing, structural graph construction, LLM-generated entity explanations, and embedding from a lightweight online answering stage.
• Score: 0.0
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Large language models have become central to modern coding assistants, but state-of-the-art systems such as Claude Code or Codex rely on very large, cloud-hosted models with significant computational cost and data-privacy implications. This work investigates whether a useful, fully local coding agent can be built around small open-source models by shifting the computational burden away from inference. We develop a Retrieval-Augmented Generation (RAG) system for scientific code understanding that strictly separates an expensive offline ingestion stage parsing, structural graph construction, LLM-generated entity explanations, and embedding from a lightweight online answering stage. The system is evaluated on a 100-question benchmark spanning eleven categories over the IPPL scientific codebase written in C++, with answers scored by an independent frontier model as the judge. Across seven answering models, we find that model family and retrieval quality matter more than parameter count, i.e. a 9B model achieves the highest average score (0.795), outperforming both larger models within our pipeline and the same models embedded in the Claude Code retrieval architecture. The results indicate that front-loading code understanding into a reusable, codebase-specialised vector store enables small local models to deliver grounded and repository-specific answers, making the agent well suited as a privacy-preserving development tool for in-house scientific codebases.

Related papers

• AgentFold: Closed-Loop Agentic Search for Protein Folding Model Design [27.154390712714264]
  We present AgentFold, a framework that formulates folding-model development as a closed-loop search over executable code variants.<n>Under a matched computational budget, AgentFold improves the best lDDT by 7.5% over independent Codex proposals.<n>The resulting intervention traces reveal recurring empirical design patterns.
  arXiv  Detail & Related papers  (2026-08-27T07:38:16Z)
• Recall Before Rerank: Benchmarking Deep Learning Models for Large-Scale Code-to-Code Retrieval [1.4062671086456253]
  This paper evaluates the effectiveness, efficiency, and scalability of contemporary deep learning models for first-stage recall in large-scale code-to-code search engines.<n>We present LLM-based code normalisation and query-rewriting schemes that yield significant gains in precision for lower-performing models.
  arXiv  Detail & Related papers  (2026-06-24T13:41:36Z)
• QiboAgent: a practitioner's guideline to open source assistants for Quantum Computing code development [0.632324422931024]
  We introduce QiboAgent, a reference implementation designed to serve as a practitioner's guideline for developing coding assistants in Quantum Computing.<n>We explore how lightweight, open-source Large Language Models provided with a custom workflow architecture compare.<n>We observe that this hybrid approach significantly reduces hallucination rates in code generation compared to a proprietary baseline, achieving a peak accuracy of 90.2%.
  arXiv  Detail & Related papers  (2026-03-16T17:05:06Z)
• Code Fingerprints: Disentangled Attribution of LLM-Generated Code [7.515488307576106]
  We study the problem of model-level code attribution, which aims to determine the source LLM responsible for generated code.<n>We propose the Disentangled Code Attribution Network (DCAN), which separates Source-Agnostic semantic information from Source-Specific stylistic representations.<n>We construct the first large-scale benchmark dataset comprising code generated by four widely used Large Language Models (LLMs) across four programming languages.
  arXiv  Detail & Related papers  (2026-03-04T15:58:36Z)
• From Code Foundation Models to Agents and Applications: A Practical Guide to Code Intelligence [150.3696990310269]
  Large language models (LLMs) have transformed automated software development by enabling direct translation of natural language descriptions into functional code.<n>We provide a comprehensive synthesis and practical guide (a series of analytic and probing experiments) about code LLMs.<n>We analyze the code capability of the general LLMs (GPT-4, Claude, LLaMA) and code-specialized LLMs (StarCoder, Code LLaMA, DeepSeek-Coder, and QwenCoder)
  arXiv  Detail & Related papers  (2025-11-23T17:09:34Z)
• Every Step Counts: Decoding Trajectories as Authorship Fingerprints of dLLMs [63.82840470917859]
  We show that the decoding mechanism of dLLMs can be used as a powerful tool for model attribution.<n>We propose a novel information extraction scheme called the Directed Decoding Map (DDM), which captures structural relationships between decoding steps and better reveals model-specific behaviors.
  arXiv  Detail & Related papers  (2025-10-02T06:25:10Z)
• Loong: Synthesize Long Chain-of-Thoughts at Scale through Verifiers [103.4410890572479]
  We introduce the Loong Project: an open-source framework for scalable synthetic data generation and verification.<n>LoongBench is a curated seed dataset containing 8,729 human-vetted examples across 12 domains.<n>LoongEnv is a modular synthetic data generation environment that supports multiple prompting strategies to produce new question-answer-code triples.
  arXiv  Detail & Related papers  (2025-09-03T06:42:40Z)
• Towards A Generalist Code Embedding Model Based On Massive Data Synthesis [35.04242699869519]
  We introduce textbfCodeR (underlineCode underlineRetrieval), a state-of-the-art embedding model for general-purpose code retrieval.<n>The superior performance of CodeR is built upon CodeR-Pile, a large-scale synthetic dataset constructed under the DRU principle.
  arXiv  Detail & Related papers  (2025-05-19T04:37:53Z)
• Learning to Solve and Verify: A Self-Play Framework for Code and Test Generation [69.62857948698436]
  Recent advances in large language models (LLMs) have improved their performance on coding benchmarks.<n>However, improvement is plateauing due to the exhaustion of readily available high-quality data.<n>We propose Sol-Ver, a self-play solver-verifier framework that jointly improves a single model's code and test generation capacity.
  arXiv  Detail & Related papers  (2025-02-20T18:32:19Z)
• OpenCoder: The Open Cookbook for Top-Tier Code Large Language Models [76.59316249991657]
  Large language models (LLMs) for code have become indispensable in various domains, including code generation, reasoning tasks and agent systems.<n>While open-access code LLMs are increasingly approaching the performance levels of proprietary models, high-quality code LLMs remain limited.<n>We introduce OpenCoder, a top-tier code LLM that not only achieves performance comparable to leading models but also serves as an "open cookbook" for the research community.
  arXiv  Detail & Related papers  (2024-11-07T17:47:25Z)
• CodeExp: Explanatory Code Document Generation [94.43677536210465]
  Existing code-to-text generation models produce only high-level summaries of code. We conduct a human study to identify the criteria for high-quality explanatory docstring for code. We present a multi-stage fine-tuning strategy and baseline models for the task.
  arXiv  Detail & Related papers  (2022-11-25T18:05:44Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.