1 paper across 1 session
A new benchmark of 118 ICPC problems for evaluating LLM reasoning in competitive coding, featuring realistic ICPC competition scenario, robust local evaluation, and a iterative repair metrics Refine@K
