Influence Tuning: Demoting Spurious Correlations via Instance Attribution and Instance-Driven Updates

By Xiaochuang Han, Yulia Tsvetkov. 2021

38 citations recorded by the source corpus.

Among the most critical limitations of deep learning NLP models are their lack of interpretability, and their reliance on spurious correlations. Prior work proposed various approaches to interpreting the black-box models to unveil the spurious correlations, but the research was primarily used in human-computer interaction scenarios. It still remains underexplored whether or how such model interpretations can be used to automatically"unlearn"confounding features. In this work, we propose influence tuning--a procedure that leverages model interpretations to update the model parameters towards a plausible interpretation (rather than an interpretation that relies on spurious patterns in the data) in addition to learning to predict the task labels. We show that in a controlled setup, influence tuning can help deconfounding the model from spurious patterns in data, significantly outperforming baseline methods that use adversarial training.

Subfields: Artificial Intelligence, 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 Safety, Robustness & Interpretability

Citations in the corpus

SELFEXPLAIN: A Self-Explaining Architecture for Neural Text Classifiers (2021, positive): “Global interpretability methods are useful not only to facilitate explainability, but also to detect and mitigate artifacts in data (Pezeshkpour et al., 2021; Han and Tsvetkov, 2021).”
WANLI: Worker and AI Collaboration for Natural Language Inference Dataset Creation (2022, background): “Han and Tsvetkov (2021) give another interpretation for this approach: for examples with the same label, the similarity of [CLS] token embeddings actually represents the similarity of gradient updates in the row of the final projection layer corresponding to that label.”
Tracing Knowledge in Language Models Back to the Training Data (2022, methodology): “A simple ﬁx proposed in previous work (Barshan et al., 2020; Han and Tsvetkov, 2021) is to unit-normalize the gradients, effectively replacing the dot product in Equation (2) with cosine similarity: ) We hereafter refer to I in Equation (4) as T RAC I N throughout the paper.”