ARCS: Agentic Retrieval-Augmented Code Synthesis with Iterative Refinement

2025-04-29

AI Summary

Code generation for high-performance computing (HPC) faces a fundamental trade-off among numerical accuracy, robustness, and computational efficiency. To address this, we propose an agent-driven retrieval-augmented code synthesis framework that introduces two key innovations: (1) a novel proxy-based Retrieval-Augmented Generation (RAG) mechanism, and (2) a state-action search tree formalism that models code generation as a dynamic, iterative process jointly optimizing for functional correctness and editing efficiency. Our approach employs chain-of-thought reasoning to guide retrieval and leverages real-time program execution feedback in a closed-loop refinement cycle across multiple iterations. Evaluated on GeeksforGeeks and HumanEval benchmarks, our method significantly outperforms conventional prompting techniques, achieving substantial improvements in both code generation and translation quality. The framework enables high-accuracy, scalable, and automated software development tailored to HPC domains.

Technology Category

• Natural Language Processing: Code Generation / Program Synthesis from Natural Language
• Cognitive Modeling & Cognitive Systems: Agent Architectures
• Search and Optimization: Sampling/Simulation-based Search

Application Category

• Search and Retrieval-Augmented AI: Agentic search
• Economics, Online Markets and Human Computation: Economic ramifications for generative AI infrastructure and applications
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

In supercomputing, efficient and optimized code generation is essential to leverage high-performance systems effectively. We propose Agentic Retrieval-Augmented Code Synthesis (ARCS), an advanced framework for accurate, robust, and efficient code generation, completion, and translation. ARCS integrates Retrieval-Augmented Generation (RAG) with Chain-of-Thought (CoT) reasoning to systematically break down and iteratively refine complex programming tasks. An agent-based RAG mechanism retrieves relevant code snippets, while real-time execution feedback drives the synthesis of candidate solutions. This process is formalized as a state-action search tree optimization, balancing code correctness with editing efficiency. Evaluations on the Geeks4Geeks and HumanEval benchmarks demonstrate that ARCS significantly outperforms traditional prompting methods in translation and generation quality. By enabling scalable and precise code synthesis, ARCS offers transformative potential for automating and optimizing code development in supercomputing applications, enhancing computational resource utilization.

Problem

• Efficient optimized code generation for supercomputing systems
• Accurate robust code synthesis completion and translation
• Automating code development to enhance computational resource utilization

Innovation

• Integrates RAG with CoT reasoning
• Uses agent-based RAG for code retrieval
• Optimizes via state-action search tree