Builder

LiveCodeBench

1 Tool

LiveCodeBench is a contamination-free benchmark for evaluating large language models on coding ability, created by researchers at UC Berkeley, MIT, and Cornell and first published in 2024. It continuously pulls new problems from LeetCode, AtCoder, and Codeforces contests and tags each by release date, so a model is only tested on problems that postdate its training cutoff. Beyond code generation, it also evaluates self-repair, code execution, and test-output prediction, addressing contamination and overfitting concerns in earlier static coding benchmarks.