ScaleBox: Enabling High-Fidelity and Scalable Code Verification for Large Language Models

2026-04-30

AI Summary

This work addresses the challenge that existing code sandbox systems struggle to simultaneously achieve high validation accuracy and computational efficiency under high concurrency, thereby limiting the training and evaluation of large language models on code generation tasks. To overcome this, the authors propose a high-fidelity, scalable code validation system featuring several key innovations: an automated special-case handling generator, a fine-grained parallel execution architecture across test cases, a multi-node distributed coordination strategy, and a configuration-driven, reproducible evaluation framework. Evaluated on LiveCodeBench, the proposed system substantially outperforms heuristic-matching baselines, demonstrating significant improvements in validation accuracy, training stability, and throughput efficiency.

Abstract

Code sandboxes have emerged as a critical infrastructure for advancing the coding capabilities of large language models, providing verifiable feedback for both RL training and evaluation. However, existing systems fail to provide accurate verification and efficiency under high-concurrency workloads. We present ScaleBox, a high-fidelity and scalable system designed to address these limitations in large-scale code training. ScaleBox introduces automated special-judge generation and management, fine-grained parallel execution across test cases with seamless multi-node coordination, and a configuration-driven evaluation suite for reproducible benchmarking. A series of experiments demonstrates that ScaleBox significantly enhances code verification accuracy and efficiency. Our further RLVR experiments show that ScaleBox substantially improves both performance on LiveCodeBench and training stability, significantly outperforming heuristic-matching baselines. By providing a reliable and high-throughput infrastructure, ScaleBox facilitates more effective research and development in large-scale code training.

Problem

• code verification
• large language models
• high concurrency
• scalability
• sandbox

Innovation

• code sandbox
• special-judge generation
• fine-grained parallel execution
• scalable code verification
• configuration-driven evaluation