datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
mahjong-winning-tiles
Mahjong Winning Tiles (胡啥牌) Benchmark
Introduction
Mahjong Winning Tiles is a benchmark dataset designed to evaluate the reasoning capability of large language models (LLMs). It focuses on assessing logical thinking, pattern recognition, and decision-making by challenging models to calculate possible winning tiles needed to complete a Mahjong hand of 13 tiles.
Background: Mahjong Rules
Mahjong is a traditional 4-player tile game originating from China and… See the full description on the dataset page: https://huggingface.co/datasets/sileixu/mahjong-winning-tiles.nassau-parcels-tiles