Can AI Assist in Olympiad Coding
Samuel Ren
TL;DR
This paper investigates whether AI can meaningfully assist experienced competitive programmers by handling implementation while humans supply the algorithmic strategy. It introduces a multi-model, GUI-enabled workflow that coordinates several AI systems to produce and refine code in C++, using problem text, algorithm descriptions, and reference materials, and evaluates this approach on USACO Gold problems. The key finding is that AI-assisted implementation can reduce time-to-solution by about 24% on six problems, though the study notes substantial limitations due to sample size and experimental scope. The work highlights practical implications for integrity, education, and the future design of competitive programming, suggesting that human-AI collaboration could become a standard part of coding workflows and that competitions may evolve to emphasize higher-level reasoning and human insight.
Abstract
As artificial intelligence programs have become more powerful, their capacity for problem-solving continues to increase, approaching top-level competitors in many olympiads. Continued development of models and benchmarks is important but not the focus of this paper. While further development of these models and benchmarks remains critical, the focus of this paper is different: we investigate how AI can assist human competitors in high-level coding contests. In our proposed workflow, a human expert outlines an algorithm and subsequently relies on an AI agent for the implementation details. We examine whether such human-AI collaboration can streamline the problem-solving process and improve efficiency, highlighting the unique challenges and opportunities of integrating AI into competitive programming contexts.