AlgorithmicResearchGroup/aria-repo-benchmark

ARIA Repo Benchmark The ARIA Repo Benchmark is part of the ARIA benchmark suite, a collection of closed-book benchmarks probing the ML knowledge that frontier models have internalized during training. This dataset contains 58 curated research paper implementations with metadata for evaluating whether ML experiments described in papers can be reproduced. Dataset Summary Size: 58 entries Coverage: Computer Vision, NLP, Time Series, Graph, Bioinformatics Purpose:… See the full description on the dataset page: https://huggingface.co/datasets/AlgorithmicResearchGroup/aria-repo-benchmark.