Language Models Are Poor Learners of Directional Inference
Tianyi Li, Mohammad Javad Hosseini, Sabine Weber, Mark Steedman
Abstract
We examine LMs’ competence of directional predicate entailments by supervised fine-tuning with prompts. Our analysis shows that contrary to their apparent success on standard NLI, LMs show limited ability to learn such directional inference; moreover, existing datasets fail to test directionality, and/or are infested by artefacts that can be learnt as proxy for entailments, yielding over-optimistic results. In response, we present BoOQA (Boolean Open QA), a robust multi-lingual evaluation benchmark for directional predicate entailments, extrinsic to existing training sets. On BoOQA, we establish baselines and show evidence of existing LM-prompting models being incompetent directional entailment learners, in contrast to entailment graphs, however limited by sparsity.
BibTeX
@inproceedings{li-etal-2022-language,
title = "Language Models Are Poor Learners of Directional Inference",
author = "Li, Tianyi and
Hosseini, Mohammad Javad and
Weber, Sabine and
Steedman, Mark",
editor = "Goldberg, Yoav and
Kozareva, Zornitsa and
Zhang, Yue",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2022",
month = dec,
year = "2022",
address = "Abu Dhabi, United Arab Emirates",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2022.findings-emnlp.64/",
doi = "10.18653/v1/2022.findings-emnlp.64",
pages = "903--921"
}