Guide the Learner: Controlling Product of Experts Debiasing Method Based on Token Attribution Similarities

TL;DR

This paper proposes a novel fine-tuning approach that uses token attribution similarities within a Product of Experts framework to better mitigate dataset biases and improve out-of-distribution performance in NLP tasks.

Contribution

It introduces a new bias mitigation method that considers attribution similarity, enhancing OOD robustness without sacrificing in-distribution accuracy.

Findings

Improves OOD performance on NLP benchmarks

Maintains in-distribution accuracy

Outperforms baseline bias mitigation methods

Abstract

Several proposals have been put forward in recent years for improving out-of-distribution (OOD) performance through mitigating dataset biases. A popular workaround is to train a robust model by re-weighting training examples based on a secondary biased model. Here, the underlying assumption is that the biased model resorts to shortcut features. Hence, those training examples that are correctly predicted by the biased model are flagged as being biased and are down-weighted during the training of the main model. However, assessing the importance of an instance merely based on the predictions of the biased model may be too naive. It is possible that the prediction of the main model can be derived from another decision-making process that is distinct from the behavior of the biased model. To circumvent this, we introduce a fine-tuning strategy that incorporates the similarity between the main and biased model attribution scores in a Product of Experts (PoE) loss function to further improve OOD performance. With experiments conducted on natural language inference and fact verification benchmarks, we show that our method improves OOD results while maintaining in-distribution (ID) performance.