ProBench: Benchmarking Large Language Models in Competitive Programming

TL;DR

ProBench is a comprehensive benchmark for evaluating large language models' competitive programming skills, focusing on reasoning, error diagnosis, and problem-solving across multiple models and datasets.

Contribution

The paper introduces ProBench, a new benchmark with a unified attribute system for assessing LLMs in competitive programming, including real test data and multi-dimensional evaluation.

Findings

QwQ-32B-Preview achieves the highest score of 20.93.

Specialized reasoning training improves programming performance.

Models trained with reasoning tasks outperform larger general models.

Abstract

With reasoning language models such as OpenAI-o3 and DeepSeek-R1 emerging, large language models (LLMs) have entered a new phase of development. However, existing benchmarks for coding evaluation are gradually inadequate to assess the capability of advanced LLMs in code reasoning. To bridge the gap for high-level code reasoning assessment, we propose ProBench to benchmark LLMs in competitive programming, drawing inspiration from the International Collegiate Programming Contest. ProBench collects a comprehensive set of competitive programming problems from Codeforces, Luogu, and Nowcoder platforms during the period from July to December 2024, obtaining real test results through online submissions to ensure the fairness and accuracy of the evaluation. We establish a unified problem attribute system, including difficulty grading and algorithm tagging. With carefully collected and annotated data in ProBench, we systematically assess 9 latest LLMs in competitive programming across multiple dimensions, including thought chain analysis, error type diagnosis, and reasoning depth evaluation. Experimental results show that QwQ-32B-Preview achieves the best score of 20.93 followed by DeepSeek-V3 with a score of 16.38, suggesting that models trained with specialized reasoning tasks significantly outperform general-purpose models (even larger than reasoning-oriented models) in programming. Further analysis also reveals key areas for programming capability enhancement, e.g., algorithm adaptability and reasoning sufficiency, providing important insights for the future development of reasoning models.