Can Pretrained Language Models (Yet) Reason Deductively?

Proceedings of the 17th Conference of the European Chapter of the Association for Computational Linguistics, pages 1447–1462 May 2-6, 2023 ©2023 Association for Computational Linguistics Can Pretrained Language Models (Yet) Reason Deductively? Zhangdie Yuan♢∗, Songbo Hu♠∗, Ivan Vuli´c♠, Anna Korhonen♠, Zaiqiao Meng♣♠† ♢Department of Computer Science and Technology, University of Cambridge ♠Language Technology Lab, University of Cambridge ♣School of Computing Science, University of Glasgow ♢♠{zy317,sh2091,iv250,alk23}@cam.ac.uk ♣zaiqiao.meng@glasgow.ac.uk Abstract Acquiring factual knowledge with Pretrained Language Models (PLMs) has attracted increas- ing attention, showing promising performance in many knowledge-intensive tasks. Their good performance has led the community to believe that the models do possess a modicum of rea- soning competence rather than merely memo- rising the knowledge. In this paper, we conduct a comprehensive evaluation of the learnable deductive (also known as explicit) reasoning capability of PLMs. Through a series of con- trolled experiments, we posit two main findings. (i) PLMs inadequately generalise learned logic rules and perform inconsistently against simple adversarial surface form edits. (ii) While the deductive reasoning fine-tuning of PLMs does improve their performance on reasoning over unseen knowledge facts, it results in catastroph- ically forgetting the previously learnt knowl- edge. Our main results suggest that PLMs can- not yet perform reliable deductive reasoning, demonstrating the importance of controlled ex- aminations and probing of PLMs’ deductive reasoning abilities; we reach beyond (mislead- ing) task performance, revealing that PLMs are still far from robust reasoning capabilities, even for simple deductive tasks. 1 Introduction Pretrained Language Models (PLMs) such as BERT (Devlin et al., 2019) and RoBERTa (Liu et al., 2019) have orchestrated tremendous progress in NLP across a large variety of downstream ap- plications. For knowledge-intensive tasks in par- ticular, these large-scale PLMs are surprisingly good at memorising factual knowledge presented in pretraining corpora (Petroni et al., 2019; Jiang et al., 2020b) and infusing knowledge from exter- nal sources (Wang et al., 2021a; Zhou et al., 2022, among others), demonstrating their effectiveness in learning and capturing knowledge. ∗Indicates equal contribution. †Corresponding author. BERT Training A raven can fly. Original bird bird bird Prediction Inference A bird can fly. A raven is a bird. A [MASK] can fly. A [MASK] can fly. A [MASK] cannot fly. The [MASK] species is decreasing. R-BERT BERT raven raven raven R-BERT Negation Query Neutral (a) (b) (1) (2) (3) (1) (2) (3) (1) (2) (3) Figure 1: Training and inference for deductive reason- ing. Given the explicit premises (a), the input BERT model is trained to get transformed into a reasoner R- BERT model by deductively predicting a previously unseen conclusion (b). This inference process requires R-BERT to understand factual knowledge and interpret rules (e.g. taxonomic relations), intervening directly in the deduction process. Automatic reasoning, a systematic process of de- riving previously unknown conclusions from given formal representations of knowledge (Lenat et al., 1990; Newell and Simon, 1956), has been a long- standing goal of AI research. In the NLP commu- nity, a modern view of this problem (Clark et al., 2020), where the formal representations …

Content truncated to 500 words