Are Red Roses Red? Evaluating Consistency of Question-Answering Models

By Marco Tulio Ribeiro, Carlos Guestrin, Sameer Singh. 2019

120 citations recorded by the source corpus.

Although current evaluation of question-answering systems treats predictions in isolation, we need to consider the relationship between predictions to measure true understanding. A model should be penalized for answering “no” to “Is the rose red?” if it answers “red” to “What color is the rose?”. We propose a method to automatically extract such implications for instances from two QA datasets, VQA and SQuAD, which we then use to evaluate the consistency of models. Human evaluation shows these generated implications are well formed and valid. Consistency evaluation provides crucial insights into gaps in existing models, while retraining with implication-augmented data improves consistency on both synthetic and human-generated implications.

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

Tailor: Generating and Perturbing Text with Semantic Controls (2021, methodology): “Because the training objective of our generator does not provide deterministic signal for (Gardner et al., 2020; Ribeiro et al., 2019). noun attachment outputs, we do not expect our verb → noun strategy to always result in generations with noun attachment.”
Evaluating Models’ Local Decision Boundaries via Contrast Sets (2020, methodology): “Automatic Construction of Contrast Sets Automatic perturbations, such as paraphrasing with back-translation or applying word replacement rules, can ﬁll in some parts of the gaps around a pivot (e.g., Ribeiro et al., 2018b, 2019).”
UNQOVERing Stereotypical Biases via Underspecified Questions (2020, methodology): “For attribute errors ( (cid:15) ), both QA models and masked LMs perform poorly due to the generally observed inconsistency in models (e.g., Ribeiro et al., 2019).”
Evaluating NLP Models via Contrast Sets (2020, methodology): “Automatic Construction of Contrast Sets Automatic perturbations, such as paraphrasing with back-translation or applying word replacement rules, can fill in some parts of the gaps around a pivot (e.g., Ribeiro et al., 2018a, 2019).”
What Do Models Learn from Question Answering Datasets? (2020, background): “Previous work has found that simple heuristics can give good performance on SQuAD 1.1 (Weissenborn et al., 2017), and successful SQuAD models lack robustness by giving inconsistent answers (Ribeiro et al., 2019) or being vulnerable to adversarial attacks (Jia and Liang, 2017; Wallace et al., 2019).”
Memory-Based Model Editing at Scale (2022, methodology): “To generate entailed questions, we use the codebase at https://github.com/marcotcr/qa consistency (Ribeiro et al., 2019), passing the question as both question and context to the entailed question generator.”
Enriching a Model's Notion of Belief using a Persistent Memory (2021, background): “Prior work on reducing inconsistency has focused on retraining the model itself to be more consistent in its answers, e.g., (Ribeiro et al., 2019; Li et al., 2019), but with imperfect results.”
Explanation-Based Human Debugging of NLP Models: A Survey (2021, background): “However, we suggest that there are several tasks which are prone to cause model bugs, not difficult to experiment on (even with lay people), but underexplored in the EBHD setting, including reading comprehension (Jia and Liang, 2017), question answering (Ribeiro et al., 2019), and natural language i”
Comprehensive Multi-Dataset Evaluation of Reading Comprehension (2019, methodology): “Several recent papers have proposed question transformations that result in out-of-distribution test examples, helping to judge the generalization capability of reading models (Ribeiro et al., 2018, 2019; Zhu et al., 2019).”
BeliefBank: Adding Memory to a Pre-Trained Language Model for a Systematic Notion of Belief (2021, background): “Most significantly, we find that improving consistency in this way improves accuracy, while earlier finetuning-based approaches report either no accuracy gains (Ribeiro et al., 2019; Minervini and Riedel, 2018; Li et al., 2019) or only slight gains (Asai and Hajishirzi, 2020).”
Beyond Leaderboards: A survey of methods for revealing weaknesses in Natural Language Inference data and models (2020, methodology): “Examples include appending irrelevant information (Jia and Liang, 2017), illustrated in Figure 2, or paraphrasing (Ribeiro et al., 2019).”
Perturbation Augmentation for Fairer NLP (2022, background): “Counterfactual data augmentation approaches are heavily inspired by recent work in the domain of robustness (Ribeiro et al., 2019; Morris et al., 2020; Khashabi et al., 2020; Balashankar et al., 2021; Liu et al., 2021; Zhang et al., 2021; Sen et al., 2021; Papakipos and Bitton, 2022).”