Abstract: Incorporating user-defined heuristic rules into neural text inference methods has the potential to align models with user intentions and domain knowledge, thereby improving interpretability.
Some results have been hidden because they may be inaccessible to you
Show inaccessible results
Feedback