Learning to Model and Ignore Dataset Bias with Mixed Capacity Ensembles
- URL: http://arxiv.org/abs/2011.03856v1
- Date: Sat, 7 Nov 2020 22:20:03 GMT
- Title: Learning to Model and Ignore Dataset Bias with Mixed Capacity Ensembles
- Authors: Christopher Clark, Mark Yatskar, and Luke Zettlemoyer
- Abstract summary: We propose a method that can automatically detect and ignore dataset-specific patterns, which we call dataset biases.
Our method trains a lower capacity model in an ensemble with a higher capacity model.
We show improvement in all settings, including a 10 point gain on the visual question answering dataset.
- Score: 66.15398165275926
- License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
- Abstract: Many datasets have been shown to contain incidental correlations created by
idiosyncrasies in the data collection process. For example, sentence entailment
datasets can have spurious word-class correlations if nearly all contradiction
sentences contain the word "not", and image recognition datasets can have
tell-tale object-background correlations if dogs are always indoors. In this
paper, we propose a method that can automatically detect and ignore these kinds
of dataset-specific patterns, which we call dataset biases. Our method trains a
lower capacity model in an ensemble with a higher capacity model. During
training, the lower capacity model learns to capture relatively shallow
correlations, which we hypothesize are likely to reflect dataset bias. This
frees the higher capacity model to focus on patterns that should generalize
better. We ensure the models learn non-overlapping approaches by introducing a
novel method to make them conditionally independent. Importantly, our approach
does not require the bias to be known in advance. We evaluate performance on
synthetic datasets, and four datasets built to penalize models that exploit
known biases on textual entailment, visual question answering, and image
recognition tasks. We show improvement in all settings, including a 10 point
gain on the visual question answering dataset.
Related papers
- ConceptDrift: Uncovering Biases through the Lens of Foundation Models [5.025665239455297]
We show that ConceptDrift can be scaled to automatically identify biases in datasets without human prior knowledge.
We propose two bias identification evaluation protocols to fill the gap in the previous work and show that our method significantly improves over SoTA methods.
Our method is not bounded to a single modality, and we empirically validate it both on image (Waterbirds, CelebA, ImageNet, and text datasets (CivilComments))
arXiv Detail & Related papers (2024-10-24T17:59:16Z) - Common-Sense Bias Discovery and Mitigation for Classification Tasks [16.8259488742528]
We propose a framework to extract feature clusters in a dataset based on image descriptions.
The analyzed features and correlations are human-interpretable, so we name the method Common-Sense Bias Discovery (CSBD)
Experiments show that our method discovers novel biases on multiple classification tasks for two benchmark image datasets.
arXiv Detail & Related papers (2024-01-24T03:56:07Z) - Stubborn Lexical Bias in Data and Models [50.79738900885665]
We use a new statistical method to examine whether spurious patterns in data appear in models trained on the data.
We apply an optimization approach to *reweight* the training data, reducing thousands of spurious correlations.
Surprisingly, though this method can successfully reduce lexical biases in the training data, we still find strong evidence of corresponding bias in the trained models.
arXiv Detail & Related papers (2023-06-03T20:12:27Z) - Debiasing Vision-Language Models via Biased Prompts [79.04467131711775]
We propose a general approach for debiasing vision-language foundation models by projecting out biased directions in the text embedding.
We show that debiasing only the text embedding with a calibrated projection matrix suffices to yield robust classifiers and fair generative models.
arXiv Detail & Related papers (2023-01-31T20:09:33Z) - Certifying Robustness to Programmable Data Bias in Decision Trees [12.060443368097102]
We certify that models produced by a learning learner are pointwise-robust to potential dataset biases.
Our approach allows specifying bias models across a variety of dimensions.
We evaluate our approach on datasets commonly used in the fairness literature.
arXiv Detail & Related papers (2021-10-08T20:15:17Z) - Learning from others' mistakes: Avoiding dataset biases without modeling
them [111.17078939377313]
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 task.
Previous work has demonstrated effective methods to circumvent these issues when knowledge of the bias is available.
We show a method for training models that learn to ignore these problematic correlations.
arXiv Detail & Related papers (2020-12-02T16:10:54Z) - Towards Robustifying NLI Models Against Lexical Dataset Biases [94.79704960296108]
This paper explores both data-level and model-level debiasing methods to robustify models against lexical dataset biases.
First, we debias the dataset through data augmentation and enhancement, but show that the model bias cannot be fully removed via this method.
The second approach employs a bag-of-words sub-model to capture the features that are likely to exploit the bias and prevents the original model from learning these biased features.
arXiv Detail & Related papers (2020-05-10T17:56:10Z) - Evaluating Models' Local Decision Boundaries via Contrast Sets [119.38387782979474]
We propose a new annotation paradigm for NLP that helps to close systematic gaps in the test data.
We demonstrate the efficacy of contrast sets by creating them for 10 diverse NLP datasets.
Although our contrast sets are not explicitly adversarial, model performance is significantly lower on them than on the original test sets.
arXiv Detail & Related papers (2020-04-06T14:47:18Z)
This list is automatically generated from the titles and abstracts of the papers in this site.
This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.