Birds Have Four Legs?! NumerSense: Probing Numerical Commonsense Knowledge of Pre-trained Language Models

By Bill Yuchen Lin, Seyeon Lee, Rahul Khanna, Xiang Ren. 2020

169 citations recorded by the source corpus.

Recent works show that pre-trained language models (PTLMs), such as BERT, possess certain commonsense and factual knowledge. They suggest that it is promising to use PTLMs as "neural knowledge bases" via predicting masked words. Surprisingly, we find that this may not work for numerical commonsense knowledge (e.g., a bird usually has two legs). In this paper, we investigate whether and to what extent we can induce numerical commonsense knowledge from PTLMs as well as the robustness of this process. To study this, we introduce a novel probing task with a diagnostic dataset, NumerSense, containing 13.6k masked-word-prediction probes (10.5k for fine-tuning and 3.1k for testing). Our analysis reveals that: (1) BERT and its stronger variant RoBERTa perform poorly on the diagnostic dataset prior to any fine-tuning; (2) fine-tuning with distant supervision brings some improvement; (3) the best supervised model still performs poorly as compared to human performance (54.06% vs 96.3% in accuracy).

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Large Language Models, Evaluation & Benchmarks

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

In Large Language Models

In Evaluation & Benchmarks

Citations in the corpus

Selection-Inference: Exploiting Large Language Models for Interpretable Logical Reasoning (2022, methodology): “Multiple choice normalisation We evaluated whether normalising log probability of the choices under the model by the token length resulted in better accuracy to avoid potential bias as reported in (Lin et al., 2020).”
CrossFit: A Few-shot Learning Challenge for Cross-task Generalization in NLP (2021, background): “…McCreery et al. 2020 mocha other/regression Chen et al. 2020a multi_news cg/summarization Fabbri et al. 2019 numer_sense qa/closed-book qa Lin et al. 2020a onestop_english cls/other Vajjala and Lučić 2018 openbookqa qa/multiple-choice qa Mihaylov et al. 2018 paws cls/paraphrase Zhang…”
Impact of Pretraining Term Frequencies on Few-Shot Reasoning (2022, background): “Prompting Prompting has been widely applied to study the factual (Petroni et al., 2019), commonsense (Davison et al., 2019; Weir et al., 2020; Lin et al., 2020), mathematical (Saxton et al., 2019), and other NLP task-related (Rad-ford et al., 2019; Shin et al., 2020) knowledge LMs acquire during…”
Investigating the Limitations of Transformers with Simple Arithmetic Tasks (2021, background): “Recent studies have explored the numerical capabilities learned by neural networks trained on large amounts of texts (Talmor et al., 2019; Jiang et al., 2019; Naik et al., 2019; Wallace et al., 2019; Lin et al., 2020; Johnson et al., 2020; Mishra et al., 2020).”
CREAK: A Dataset for Commonsense Reasoning over Entity Knowledge (2021, background): “Another recent commonsense-focused dataset [Lin et al., 2020], focuses on probing numeric claims.”
Rainier: Reinforced Knowledge Introspector for Commonsense Question Answering (2022, methodology): “The knowledge’s domain distribution is strongly tied to the domain of the benchmark (e.g. scientiﬁc for QASC and QuaRTz, social for SocialIQA and Winogrande, numerical for Nu-merSense).”
Do Language Embeddings capture Scales? (2020, background): “, and numerical common sense (Lin et al., 2020).”
Knowledge-Augmented Methods for Natural Language Processing (2023, methodology): “Apart from the benchmarking and modeling, we also introduce the analysis works that aim to provide a deeper understanding the common-sense knowledge of pre-trained LMs: LAMA Probing* (Petroni et al., 2019), NumerSense (Lin et al., 2020a), and RICA* (Zhou et al., 2020).”
Investigating the Limitations of the Transformers with Simple Arithmetic Tasks (2021, unknown): “Recent studies have explored the numerical capabilities learned by neural networks trained on large amounts of texts (Talmor et al., 2019; Jiang et al., 2019; Naik et al., 2019; Wallace et al., 2019; Lin et al., 2020; Johnson et al., 2020; Mishra et al., 2020).”
Adaptive Chameleon or Stubborn Sloth: Revealing the Behavior of Large Language Models in Knowledge Conflicts (2023, background): “Many past studies have explored the elicitation of parametric memory in language models, such as commonsense or factual knowledge probing (Petroni et al., 2019; Lin et al., 2020; Zhang et al., 2021; West et al., 2022; Chen et al., 2023; Wang et al., 2023).”
Do Language Embeddings capture Scales? (2020, unknown): “, and numerical common sense (Lin et al., 2020).”
Reasoning with Transformer-based Models: Deep Learning, but Shallow Reasoning (2021, result): “However, several studies [Forbes et al., 2019, Zhou et al., 2020b, Lin et al., 2020, Boratko et al., 2020, Singh et al., 2021] show that when the datasets are speciﬁcally changed to target the weaknesses of transformer-based models (for example, by adversarial instances), the models fail.”
Generated Knowledge Prompting for Commonsense Reasoning (2021, methodology): “NumerSense (Lin et al., 2020) consists of numerical statements about common objects and concepts where for each sentence we need to recover a masked number word.”
Probing Toxic Content in Large Pre-Trained Language Models (2021, background): “Previous work on probing PTLMs focuses on their syntactic and semantic limitations (Hewitt and Manning, 2019; Marvin and Linzen, 2018), lack of domain-speciﬁc knowledge (Jin et al., 2019), and absence of commonsense (Petroni et al., 2019; Lin et al., 2020).”
Representing Numbers in NLP: a Survey and a Vision (2021, background): “Two recent datasets for numeric commonsense facts are Numbergame (Mishra et al., 2020) and NumerSense (Lin et al., 2020).”
Natural Language Processing with Commonsense Knowledge: A Survey (2021, background): “• N UMER S ENSE [74] explores the limitations of LLMs like BERT and RoBERTa in understanding and recalling numerical commonsense knowledge.”