Learning What Makes a Difference from Counterfactual Examples and Gradient Supervision

By Damien Teney, Ehsan Abbasnejad, Anton van den Hengel. 2020

127 citations recorded by the source corpus.

One of the primary challenges limiting the applicability of deep learning is its susceptibility to learning spurious correlations rather than the underlying mechanisms of the task of interest. The resulting failure to generalise cannot be addressed by simply using more data from the same distribution. We propose an auxiliary training objective that improves the generalization capabilities of neural networks by leveraging an overlooked supervisory signal found in existing datasets. We use pairs of minimally-different examples with different labels, a.k.a counterfactual or contrasting examples, which provide a signal indicative of the underlying causal structure of the task. We show that such pairs can be identified in a number of existing datasets in computer vision (visual question answering, multi-label image classification) and natural language processing (sentiment analysis, natural language inference). The new training objective orients the gradient of a model's decision function with pairs of counterfactual examples. Models trained with this technique demonstrate improved performance on out-of-distribution test sets.

Subfields: Natural Language Processing, Representation & Self-Supervised Learning

How researchers see this paper

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

In Natural Language Processing

In Representation & Self-Supervised Learning

Citations in the corpus

Tailor: Generating and Perturbing Text with Semantic Controls (2021, background): “It has been widely applied to a variety of tasks, e.g., changing text style (Reid and Zhong, 2021), mitigating dataset biases (Gardner et al., 2021), explaining model behaviors (Ross et al., 2021), and improving model generalization (Teney et al., 2020; Wu et al., 2021).”
Polyjuice: Generating Counterfactuals for Explaining, Evaluating, and Improving Models (2021, background): “…on tuning the amount and the distribution of counterfactual augmentation, as well as other ways of incorporating counterfactuals, such as having explicit terms in the loss function for contrasting counterfactuals with original data (Teney et al., 2020), or other forms of contrastive learning.”
Data and its (dis)contents: A survey of dataset development and use in machine learning research (2020, background): “Teney et al.(85) propose an auxiliary training objective using counterfactually labeled data to guide models toward better decision boundaries.”
Learning to Imagine: Integrating Counterfactual Thinking in Neural Discrete Reasoning (2022, background): “…effect of counterfactual samples by combining with different learning paradigms, such as adversarial training (Zhu et al., 2020; Fu et al., 2020; Teney et al., 2020), contrastive learning (Liang et al., 2020), causal graph (Gokhale et al., 2020), posterior regularization (Ramakrishnan et al.,…”
LPF: A Language-Prior Feedback Objective Function for De-biased Visual Question Answering (2021, methodology): “Besides, counterfactual data augmentation techniques [1, 9, 22, 30, 34] are proposed to balance the training data and boost the generalization performance of the model.”
A Closer Look at the Robustness of Vision-and-Language Pre-trained Models (2020, background): “Considerable effort [3, 29, 33, 48, 53, 1] has been invested on VQA-CP along 3 dimensions: (i) compensating for question-answer distribution patterns through a regularizer based on an auxiliary model [48, 8, 14, 67, 21, 29]; (ii) taking advantage of additional supervision from human-generated attent”
Counterfactual Generative Networks (2021, background): “, 2020) and visual question answering (Teney et al., 2020).”
Explaining The Efficacy of Counterfactually-Augmented Data (2020, background): “In complex domains with high dimensional data, such as computer vision or natural language processing (NLP), what precisely are the relevant variables and what graph relates them? One recent line of work (Kaushik et al., 2020; Teney et al., 2020; Srivastava et al., 2020) has sought to inject causal ”
Explaining NLP Models via Minimal Contrastive Editing (MiCE) (2020, background): “Recent work explores the use of minimal edits changing true labels for evaluation (Gardner et al., 2020) and data augmentation (Kaushik et al., 2020; Teney et al., 2020), whereas we focus on minimal edits changing model predictions for explanation .”