pyrrosk/PetriBench

PetriBenchV1 PetriBench evaluates language-model reasoning over dynamic state spaces using procedurally generated Petri nets. PetriBenchV1 contains 4,800 questions across six reasoning tasks and three difficulty levels. Evaluation data only. Do not use this benchmark for model training, copy it into another corpus, mirror it, or re-upload it elsewhere. Direct other researchers to this gated repository instead. Access This repository is gated so that users… See the full description on the dataset page: https://huggingface.co/datasets/pyrrosk/PetriBench.