EffiCodeBench is a runtime-supervised benchmark for evaluating whether models can judge code execution efficiency. Each instance compares two functionally equivalent implementations of the same programming task. The implementation with the lower measured runtime is treated as the preferred side.
The dataset is designed for CodeEffiJudge-style pairwise efficiency judgment: given two candidate programs, a model should decide which side is faster or whether a… See the full description on the dataset page:
https://huggingface.co/datasets/JinNian0072/efficodebench-dataset.
