Large NLP models like BERT can answer questions about a document thanks to the transformer network, a sequence-processing architecture that retains information across much longer sequences than previous methods. But transformers have had little success in reinforcement learning — until now.

Summarizing a document using original words is a longstanding problem for natural language processing. Researchers recently took a step toward human-level performance in this task, known as abstractive summarization, as opposed to extractive summarization.

Drugs undergo rigorous experimentation and clinical trials to gain regulatory approval, while dietary supplements get less scrutiny. Even when a drug study reveals an interaction with supplements, the discovery tends to receive little attention.

Despite dramatic recent progress, natural language generation remains an iffy proposition. Even users of the muscular GPT-2 text generator have to press the button a number of times to get sensible output. But researchers are figuring out how to exert greater control over generated text.

Watson set a high bar for language understanding in 2011, when it famously whipped human competitors in the televised trivia game show Jeopardy! IBM’s special-purpose AI required around $1 billion. Research suggests that today’s best language models can accomplish similar tasks right off the shelf.

OpenAI raised eyebrows in February when it announced — and withheld — the full version of its groundbreaking language model, GPT-2. Six months later, the company has re-examined the decision.

Less than a month after XLNet overtook BERT, the pole position in natural language understanding changed hands again. RoBERTa is an improved BERT pretraining recipe that beats its forbear, becoming the new state-of-the-art language model — for the moment.