SEP 10, 2026 · PREPRINT
E-CONAN (Entailment, CONtradition And Neutral) Benchmarks: Arabic Textual Entailment and Natural Inference Datasets
arXiv
A new benchmark dataset paper without peer review, presenting resource development and descriptive evaluation across multiple models without controlled comparisons or definitive performance claims.
Study details
DesignBenchmark dataset development with descript…
InterventionE-CONAN benchmark datasets (E-CONAN-2 for 2…
ComparatorComparison datasets: ArNLI and XNLI; baseli…