Featured Demos

Macaw is a high-performance question-answering (QA) model capable of outperforming other popular current language models, all while being an order of magnitude smaller. This demo allows you to explore Macaw's answers and compare them to those of the popular GPT-3 language model on a benchmark set of questions.

Delphi is intended to demonstrate what state-of-the-art models can accomplish today on machine ethics as well as to highlight their limitations.

Like RuleTaker, ProofWriter determines whether statements are True or False based on rules given in natural language - but also generates the proof of its answers.

This is an experimental prototype that aims to render scientific papers in HTML so they can be more easily read by screen readers or on mobile devices.

ModularQA is a neuro-symbolic question-answering system that answers complex questions by asking a series of sub-questions to existing simpler QA systems or symbolic modules. It explains each of its reasoning steps in language, in terms of a simple question and its answer as produced by a simpler model or a math… more

This work focuses specifically on identifying biases in question answering (QA) models. If these models are blindly deployed in real-life settings, the biases within these models could cause real harm, which raises the question; how extensive are social stereotypes in question-answering models?

In new joint work at AI2 and UW, we study how often popular NLP components produce problematic content, what might trigger this neural toxic degeneration from a given system, and whether or not it can be successfully avoided. We also study how much toxicity is present in the web text that these systems learned… more

Our fact verification demo was built using the SciFact dataset, a collection of 1.4K expert-written scientific claims paired with evidence-containing abstracts, and annotated with labels and rationales.

UnifiedQA is a single pre-trained QA model that performs surprisingly well across 17 QA datasets spanning 4 diverse formats. Fine-tuning UnifiedQA into specialized models results in a new state-of-the-art on 6 datasets, establishing this model as a strong starting point for building QA systems.