Breaking NLI Systems with Sentences that Require Simple Lexical Inferences

By Max Glockner, Vered Shwartz, Yoav Goldberg. 2018

391 citations recorded by the source corpus.

We create a new NLI test set that shows the deficiency of state-of-the-art models in inferences that require lexical and world knowledge. The new examples are simpler than the SNLI test set, containing sentences that differ by at most one word from sentences in the training set. Yet, the performance on the new test set is substantially worse across systems trained on SNLI, demonstrating that these systems are limited in their generalization ability, failing to capture many simple inferences.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Natural Language Processing, 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 Natural Language Processing

In Evaluation & Benchmarks

Citations in the corpus

SWAG: A Large-Scale Adversarial Dataset for Grounded Commonsense Inference (2018, background): “As a result, models trained on such datasets with hu-1 man biases run the risk of over-estimating the actual performance on the underlying task, and are vulnerable to adversarial or out-of-domain examples (Wang et al., 2018; Glockner et al., 2018).”
Experience Grounds Language (2020, methodology): “This disconnect is driven home by studies of bias (Gururangan et al., 2018; Glockner et al., 2018) and techniques like Adversarial Filtering (Zellers et al., 2018, 2019b), which elucidate the biases models exploit in lieu of understanding.”
Adversarial NLI: A New Benchmark for Natural Language Understanding (2019, background): “NLI systems can often be broken merely by performing simple lexical substitutions (Glockner et al., 2018), and struggle with quanti-ﬁers (Geiger et al., 2018) and certain superﬁcial syntactic properties (McCoy et al., 2019).”
Probing Neural Network Comprehension of Natural Language Arguments (2019, background): “There is much recent work focusing on statistical cues in datasets in vision (Jo and Bengio, 2017) and NLP (Sanchez et al., 2018; McCoy et al., 2019; Gururangan et al., 2018; Glockner et al., 2018; Poliak et al., 2018; Rajpurkar et al., 2018; Jia and Liang, 2017).”
HellaSwag: Can a Machine Really Finish Your Sentence? (2019, neutral): “This contrasts with past work on adversarial examples (e.g. Jia and Liang, 2017; Glockner et al., 2018; Belinkov and Bisk, 2018) which consider cases where an out-of-distribution test set is constructed to be adversarial.”
BoolQ: Exploring the Surprising Difficulty of Natural Yes/No Questions (2019, background): “The success of MultiNLI might also be surprising given recent concerns about the generalization abilities of models trained on it (Glockner et al., 2018), particularly related to “annotation artifacts” caused by using crowd workers to write the hypothesis statements (Gururangan et al., 2018).”
Right for the Wrong Reasons: Diagnosing Syntactic Heuristics in Natural Language Inference (2019, background): “Some of these works focus on speciﬁc semantic phenomena, such as lexical inferences (Glockner et al., 2018) and quantiﬁers (Geiger et al., 2018), or biases based on speciﬁc words (Sanchez et al., 2018).”
Multi-Task Deep Neural Networks for Natural Language Understanding (2019, background): “At last, we also would like to verify whether MT-DNN is resilience against adversarial attacks (Glockner et al., 2018; Talman and Chatzikyriakidis, 2018; Liu et al., 2019).”
HateCheck: Functional Tests for Hate Speech Detection Models (2020, methodology): “In its use of contrastive cases for model evaluation, H ATE C HECK builds on a long history of minimally-contrastive pairs in NLP (e.g. Levesque et al., 2012; Sennrich, 2017; Glockner et al., 2018; Warstadt et al., 2020).”
Shortcut learning in deep neural networks (2020, background): “Feature combination crucially depends on shortcut features like word length [59, 60, 14, 61], and consequently leads to a severe lack of robustness such as an inability to generalise to more challenging test conditions [62, 63, 64, 65].”
Open-Domain Conversational Agents: Current Progress, Open Problems, and Future Directions (2020, background): “Good progress is being made on these tasks, although questions still remain about how much true generalization is actually occurring (Glockner, Shwartz, and Goldberg, 2018; Gururangan et al., 2018).”
e-SNLI: Natural Language Inference with Natural Language Explanations (2018, background): “Glockner et al. [11] introduce a toy dataset, BreakingNLI, to test whether natural language inference models capture world knowledge and generalize beyond statistical regularities.”
WANLI: Worker and AI Collaboration for Natural Language Inference Dataset Creation (2022, background): “For crowdsourcing NLI datasets, where the anno-tator is given a premise and asked to write a hypothesis of each label (Bowman et al., 2015; Williams et al., 2018), the presence of annotation artifacts is especially well-studied (Gururangan et al., 2018; McCoy et al., 2019; Glockner et al., 2018).”
A Survey of Data Augmentation Approaches for NLP (2021, methodology): “Kang et al. (2018) and Glockner et al. (2018) create AVEs for textual entailment using WordNet relations.”
Polyjuice: Generating Counterfactuals for Explaining, Evaluating, and Improving Models (2021, methodology): “…2020), likely because their distributions are less similar to the original SST-2 training data. challenge sets (Kim et al., 2019; Naik et al., 2018; Glockner et al., 2018; Wang et al., 2019); it exhibits comparable (or better) gains than m-CAD (manual counterfactuals) with less implementation and…”
Analysis Methods in Neural Language Processing: A Survey (2018, background): “For instance, Sanchez et al. (2018) and Glockner et al. (2018) extracted examples from SNLI (Bow-man et al., 2015) and replaced speciﬁc words such as hypernyms, synonyms, and antonyms, followed by manual veriﬁcation.”
Learning the Difference that Makes a Difference with Counterfactually-Augmented Data (2019, background): “[6] demonstrate that simply replacing words by synonyms or hypernyms, which should not alter the applicable label, nevertheless breaks ML-based NLI systems.”
Dynabench: Rethinking Benchmarking in NLP (2021, background): “…for social (Rudinger et al., 2018; May et al., 2019; Nangia et al., 2020) and/or linguistic competencies (Geiger et al., 2018; Naik et al., 2018; Glockner et al., 2018; White et al., 2018; Warstadt et al., 2019; Gauthier et al., 2020; Hossain et al., 2020; Jeretic et al., 2020; Lewis et al.,…”
Are We Modeling the Task or the Annotator? An Investigation of Annotator Bias in Natural Language Understanding Datasets (2019, neutral): “In this paper, we continue recent efforts to understand biases that are introduced during the process of data creation (Levy et al., 2015; Schwartz et al., 2017; Gururangan et al., 2018; Glockner et al., 2018; Poliak et al., 2018; Tsuchiya, 2018; Aharoni and Goldberg, 2018; Paun et al., 2018).”
Evaluating Models’ Local Decision Boundaries via Contrast Sets (2020, background): “The idea of creating challenging contrastive evaluation sets has a long history (Levesque et al., 2011; Ettinger et al., 2017; Glockner et al., 2018; Naik et al., 2018; Isabelle et al., 2017).”