WillowNLtoFOL is a high-quality, structurally diverse dataset designed for evaluating the compositional generalization capabilities of neural semantic parsing models on the Natural Language to First-Order Logic (NL-to-FOL) translation task.
Translating natural language into formal logic requires nuanced linguistic understanding and the systematic recombination of learned structures. Existing datasets often suffer… See the full description on the dataset page: https://huggingface.co/datasets/iedeveci/WillowNLtoFOL.
