Competitive programming has emerged as a critical benchmark for evaluating the reasoning and coding capabilities of Large Language Models (LLMs). Despite impressive progress on existing benchmarks, we argue that current evaluations overstate model proficiency, masking a substantial gap between LLMs and elite human programmers. This gap arises from two key limitations: insufficient difficulty and scope of benchmark problems, and evaluation bias from low-quality test cases. To address these shortcomings, we present AetherCode, a new benchmark that draws problems from premier programming competitions such as IOI and ICPC, offering broader coverage and higher difficulty. AetherCode further incorporates comprehensive, expert-validated test suites built through a hybrid of automated generation and human curation, ensuring rigorous and reliable assessment. By combining challenging problem design with robust evaluation, AetherCode provides a more faithful measure of LLM capabilities and sets a new standard for future research in code reasoning.
@article{arxiv.2508.16402,
title = {AetherCode: Evaluating LLMs' Ability to Win In Premier Programming Competitions},
author = {Zihan Wang and Jiaze Chen and Zhicheng Liu and Markus Mak and Yidi Du and Geonsik Moon and Luoqi Xu and Aaron Tua and Kunshuo Peng and Jiayi Lu and Mingfei Xia and Boqian Zou and Chenyang Ran and Guang Tian and Shoutai Zhu and Yeheng Duan and Zhenghui Kang and Zhenxing Lin and Shangshu Li and Qiang Luo and Qingshen Long and Zhiyong Chen and Yihan Xiao and Yurong Wu and Daoguang Zan and Yuyi Fu and Mingxuan Wang and Ming Ding},
journal= {arXiv preprint arXiv:2508.16402},
year = {2025}
}