Supervising Model Attention with Human Explanations for Robust Natural Language Inference

By Joe Stacey, Yonatan Belinkov, Marek Rei. 2021

53 citations recorded by the source corpus.

Natural Language Inference (NLI) models are known to learn from biases and artefacts within their training data, impacting how well they generalise to other unseen datasets. Existing de-biasing approaches focus on preventing the models from learning these biases, which can result in restrictive models and lower performance. We instead investigate teaching the model how a human would approach the NLI task, in order to learn features that will generalise better to previously unseen examples. Using natural language explanations, we supervise the model’s attention weights to encourage more attention to be paid to the words present in the explanations, significantly improving model performance. Our experiments show that the in-distribution improvements of this method are also accompanied by out-of-distribution improvements, with the supervised models learning from features that generalise better to other NLI datasets. Analysis of the model indicates that human explanations encourage increased attention on the important words, with more attention paid to words in the premise and less attention paid to punctuation and stopwords.

Subfields: Artificial Intelligence, Natural Language Processing, Safety, Robustness & Interpretability

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

In Safety, Robustness & Interpretability

Citations in the corpus

Training Language Models with Language Feedback (2022, methodology): “, 2019, inter alia) or auxiliary losses (Stacey et al., 2021) that cannot be straightforwardly generalized to use natural language feedback.”
When Can Models Learn From Explanations? A Formal Framework for Understanding the Roles of Explanation Data (2021, background): “priors over how features are weighted or extracted (Zaidan et al., 2007; Small et al., 2011; Zhang et al., 2016; Ross et al., 2017; Bao et al., 2018; Selvaraju et al., 2019; Liang et al., 2020; Stammer et al., 2020; Pruthi et al., 2021; Stacey et al., 2022).”
The Unreliability of Explanations in Few-shot Prompting for Textual Reasoning (2022, methodology): “Prior research has also explored using explanations as additional supervision to train joint models (Hancock et al., 2018; Dua et al., 2020; Lamm et al., 2021; Stacey et al., 2022).”
The Unreliability of Explanations in Few-Shot In-Context Learning (2022, methodology): “Prior research has also explored using explanations as additional supervision to train joint models that jointly predict the labels and explanations (Hancock et al., 2018; Dua et al., 2020; Lamm et al., 2021; Stacey et al., 2022).”
Training Language Models with Natural Language Feedback (2022, methodology): “Prior work explores ways of using explanations to train text classiﬁcation models, with mixed re-sults (Camburu et al., 2018; Stacey et al., 2021; Pruthi et al., 2021; Wiegreffe et al., 2021; Hase and Bansal, 2021; Lampinen et al., 2022, inter alia ).”
Learning from Natural Language Feedback (2024, methodology): “, 2019, inter alia) or auxiliary losses (Stacey et al., 2021) that cannot be straightforwardly generalized to use natural language feedback.”
Training Language Models with Language Feedback at Scale (2023, methodology): “Prior work explores ways of using explanations to train text classification models, with mixed results (Camburu et al., 2018; Stacey et al., 2021; Pruthi et al., 2021; Wiegreffe et al., 2021; Hase & Bansal, 2021; Lampinen et al., 2022, inter alia ).”
Explainability for Large Language Models: A Survey (2023, background): “Counterfactual prompts have been proposed to perturb key components of a prompt: symbols, patterns, and text [Madaan and Yazdanbakhsh 2022].”