Language models (LMs) have been shown to memorize a great deal of factual knowledge contained in their training data. But when an LM generates an assertion, it is often difficult to determine where it learned this information and whether it is true. In this paper, we propose the problem of fact tracing: identifying which training examples taught an LM to generate a particular factual assertion. Prior work on training data attribution (TDA) may offer effective tools for identifying such examples, known as “proponents”. We present the first quantitative benchmark to evaluate this. We compare two popular families of TDA methods — gradient-based and embedding-based — and find that much headroom remains. For example, both methods have lower proponent-retrieval precision than an information retrieval baseline (BM25) that does not have access to the LM at all. We identify key challenges that may be necessary for further improvement such as overcoming the problem of gradient saturation, and also show how several nuanced implementation details of existing neural TDA methods can significantly improve overall fact tracing performance.

We present the Language Interpretability Tool (LIT), an open-source platform for visualization and understanding of NLP models. We focus on core questions about model behavior: Why did my model make this prediction? When does it perform poorly? What happens under a controlled change in the input? LIT integrates local explanations, aggregate analysis, and counterfactual generation into a streamlined, browser-based interface to enable rapid exploration and error analysis. We include case studies for a diverse set of workflows, including exploring counterfactuals for sentiment analysis, measuring gender bias in coreference systems, and exploring local behavior in text generation. LIT supports a wide range of models—including classification, seq2seq, and structured prediction—and is highly extensible through a declarative, framework-agnostic API. LIT is under active development, with code and full documentation available at https://github.com/pair-code/lit.

(Bolukbasi et al., 2016) demonstrated that pretrained word embeddings can inherit gender bias from the data they were trained on. We investigate how this bias affects downstream classification tasks, using the case study of occupation classification (De-Arteaga et al., 2019). We show that traditional techniques for debiasing embeddings can actually worsen the bias of the downstream classifier by providing a less noisy channel for communicating gender information. With a relatively minor adjustment, however, we show how these same techniques can be used to simultaneously reduce bias and maintain high classification accuracy.

We design a generic framework for learning a robust text classification model that achieves high accuracy under different selection budgets (a.k.a selection rates) at test-time. We take a different approach from existing methods and learn to dynamically filter a large fraction of unimportant words by a low-complexity selector such that any high-complexity state-of-art classifier only needs to process a small fraction of text, relevant for the target task. To this end, we propose a data aggregation method to train the classifier, allowing it to achieve competitive performance on fractured sentences. On four benchmark text classification tasks, we demonstrate that the framework gains consistent speedup with little degradation in accuracy on various selection budgets.