NLP models are shown to suffer from robustness issues, i.e., a model's prediction can be easily changed under small perturbations to the input. In this work, we present a Controlled Adversarial Text Generation (CAT-Gen) model that, given an input text, generates adversarial texts through controllable attributes that are known to be irrelevant to task labels. For example, in order to attack a model for sentiment classification over product reviews, we can use the product categories as the controllable attribute which should not change the sentiment of the reviews. Experiments on real-world NLP datasets demonstrate that our method can generate more diverse and fluent adversarial texts, compared to many existing adversarial text generation approaches. We further use our generated adversarial examples to improve models through adversarial training, and we demonstrate that our generated attacks are more robust against model retraining and different model architectures. * This research was conducted during the author's internship at Google Research.
Developing robust NLP models that perform well on many, even small, slices of data is a significant but important challenge, with implications from fairness to general reliability. To this end, recent research has explored how models rely on spurious correlations, and how counterfactual data augmentation (CDA) can mitigate such issues. In this paper we study how and why modeling counterfactuals over multiple attributes can go significantly further in improving model performance. We propose RDI, a context-aware methodology which takes into account the impact of secondary attributes on the model's predictions and increases sensitivity for secondary attributes over reweighted counterfactually augmented data. By implementing RDI in the context of toxicity detection, we find that accounting for secondary attributes can significantly improve robustness, with improvements in sliced accuracy on the original dataset up to 7% compared to existing robustness methods. We also demonstrate that RDI generalizes to the coreference resolution task and provide guidelines to extend this to other tasks.
Abstract. The goal of this paper is to provide an accurate pixel-level segmentation of a deformable foreground object in an image. We combine state-of-the-art local image segmentation techniques with a global object-specific contour model to form a coherent energy function over the outline of the object and the pixels inside it. The energy function includes terms from a variant of the TextonBoost method, which labels each pixel as either foreground or background. It also includes terms over landmark points from a LOOPS model [1], which combines global object shape with landmark-specific detectors. We allow the pixel-level segmentation and object outline to inform each other through energy potentials so that they form a coherent object segmentation with globally consistent shape and appearance. We introduce an inference method to optimize this energy that proposes moves within the complex energy space based on multiple initial oversegmentations of the entire image. We show that this method achieves state-of-the-art results in precisely segmenting articulated objects in cluttered natural scenes.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.