Daniel M. Ziegler

TL;DR: GPT-3 achieves strong performance on many NLP datasets, including translation, question-answering, and cloze tasks, as well as several tasks that require on-the-fly reasoning or domain adaptation, such as unscrambling words, using a novel word in a sentence, or performing 3-digit arithmetic.

TL;DR: This article showed that scaling up language models greatly improves task-agnostic, few-shot performance, sometimes even reaching competitiveness with prior state-of-the-art fine-tuning approaches.

TL;DR: The authors use reinforcement learning to fine-tune a summarization policy according to human feedback, which results in better summaries than optimizing ROUGE according to humans, and transfer to CNN/DM news articles, producing summaries nearly as good as the human reference.

TL;DR: This paper builds on advances in generative pretraining of language models to apply reward learning to four natural language tasks: continuing text with positive sentiment or physically descriptive language, and summarization tasks on the TL;DR and CNN/Daily Mail datasets.

TL;DR: The Crash Hoare logic (CHL), which extends traditionalHoare logic with a crash condition, a recovery procedure, and logical address spaces for specifying disk states at different abstraction levels, is introduced, which reduces the proof effort for developers through proof automation.