Behavior Analysis of NLI Models: Uncovering the Influence of Three Factors on Robustness

By Vicente Iván Sánchez Carmona, Jeff Mitchell, Sebastian Riedel. 2018

45 citations recorded by the source corpus.

Natural Language Inference is a challenging task that has received substantial attention, and state-of-the-art models now achieve impressive test set performance in the form of accuracy scores. Here, we go beyond this single evaluation metric to examine robustness to semantically-valid alterations to the input data. We identify three factors - insensitivity, polarity and unseen pairs - and compare their impact on three SNLI models under a variety of conditions. Our results demonstrate a number of strengths and weaknesses in the models’ ability to generalise to new in-domain instances. In particular, while strong performance is possible on unseen hypernyms, unseen antonyms are more challenging for all the models. More generally, the models suffer from an insensitivity to certain small but semantically significant alterations, and are also often influenced by simple statistical correlations between words and training labels. Overall, we show that evaluations of NLI models can benefit from studying the influence of factors intrinsic to the models or found in the dataset used.

Subfields: Artificial Intelligence, Natural Language Processing

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 Natural Language Processing

Citations in the corpus

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).”
Right for the Wrong Reasons: Diagnosing Syntactic Heuristics in Natural Language Inference (2019, background): “On this task as well, neural models have been shown to learn shallow heuristics that are based on the distributions of speciﬁc words (Naik et al., 2018; Sanchez et al., 2018).”
Analysis Methods in Neural Language Processing: A Survey (2018, methodology): “Automatically-built datasets are much larger, ranging from several thousands to close to a hundred thousand (Sennrich, 2017), or even more than one million examples (Linzen et al., 2016).”
Make Up Your Mind! Adversarial Generation of Inconsistent Natural Language Explanations (2019, methodology): “In fact, during our scanning over explanations, we also experimented with a few manually created potential adversarial hypothesis from Carmona et al. [5].”
The Curse of Performance Instability in Analysis Datasets: Consequences, Source, and Suggestions (2020, background): “, 2018; Shwartz and Dagan, 2018; Carmona et al., 2018), some are more syntactic-related (McCoy et al., 2019b; Nie et al., 2019; Geiger et al., 2019), some also involved logical-related study (Minervini and Riedel, 2018; Wang et al., 2019). Naik et al. (2018a) proposed a suite of analysis sets coveri”
Combining Feature and Instance Attribution to Detect Artifacts (2021, background): “…trained and evaluated on such corpora tend to exploit common artifacts present in the crowdsourced annotations, e.g., premise-hypothesis pairs with overlapping to-kens and hypotheses containing negations both correlate with labels (Gururangan et al., 2018; Sanchez et al., 2018; Naik et al., 2018).”
Beyond Leaderboards: A survey of methods for revealing weaknesses in Natural Language Inference data and models (2020, background): “Adversarial Evaluation refers to generating data with the aim to “fool” a target model.”
Analyzing Compositionality-Sensitivity of NLI Models (2018, methodology): “Existing Analysis on NLI : Previous work on analyzing NLI models (Glockner, Shwartz, and Goldberg 2018; Carmona, Mitchell, and Riedel 2018) has focused on models’ limited ability in identifying lexical semantics that were rare or unseen during training.”