Fine-tuning a transformer model for textual entailment over expressive contexts generated from description logic knowledge bases. Specifically, the model is given a context (a set of facts and rules) and a question. The model should answer with "True" if the question is logically implied from the context, "False" if it contradicts the context, and "Unknown" if none of the two.

For more info please see our paper.

DELTA_M is a DeBERTaV3 large model fine-tuned on the DELTA_D dataset.

@misc{poulis2024transformers,
      title={Transformers in the Service of Description Logic-based Contexts}, 
      author={Angelos Poulis and Eleni Tsalapati and Manolis Koubarakis},
      year={2024},
      eprint={2311.08941},
      archivePrefix={arXiv},
      primaryClass={cs.CL}
}