Learning from others' mistakes: Avoiding dataset biases without modeling them

12/02/2020
by   Victor Sanh, et al.
10

State-of-the-art natural language processing (NLP) models often learn to model dataset biases and surface form correlations instead of features that target the intended underlying task. Previous work has demonstrated effective methods to circumvent these issues when knowledge of the bias is available. We consider cases where the bias issues may not be explicitly identified, and show a method for training models that learn to ignore these problematic correlations. Our approach relies on the observation that models with limited capacity primarily learn to exploit biases in the dataset. We can leverage the errors of such limited capacity models to train a more robust model in a product of experts, thus bypassing the need to hand-craft a biased model. We show the effectiveness of this method to retain improvements in out-of-distribution settings even if no particular bias is targeted by the biased model.

READ FULL TEXT
07/09/2019

On Adversarial Removal of Hypothesis-only Bias in Natural Language Inference

Popular Natural Language Inference (NLI) datasets have been shown to be ...
04/05/2022

OccamNets: Mitigating Dataset Bias by Favoring Simpler Hypotheses

Dataset bias and spurious correlations can significantly impair generali...
02/16/2022

Decorrelate Irrelevant, Purify Relevant: Overcome Textual Spurious Correlations from a Feature Perspective

Natural language understanding (NLU) models tend to rely on spurious cor...
11/07/2020

Learning to Model and Ignore Dataset Bias with Mixed Capacity Ensembles

Many datasets have been shown to contain incidental correlations created...
04/17/2021

Competency Problems: On Finding and Removing Artifacts in Language Data

Much recent work in NLP has documented dataset artifacts, bias, and spur...
03/18/2022

Pseudo Bias-Balanced Learning for Debiased Chest X-ray Classification

Deep learning models were frequently reported to learn from shortcuts li...
09/01/2021

Don't Discard All the Biased Instances: Investigating a Core Assumption in Dataset Bias Mitigation Techniques

Existing techniques for mitigating dataset bias often leverage a biased ...