Melanie Subbiah

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: It is argued that state-of-the-art large language models are susceptible to the generation of unsafe text and haveulty rejecting unsafe advice, and it is argued for further studies of safety and the assessment of commonsense physical safety in models before release.

TL;DR: This work distinguishes types of text that can lead to physical harm and establishes one particularly underexplored category: covertly unsafe text, which is further broken down with respect to the system’s information and discusses solutions to mitigate the generation of text in each of these subcategories.

TL;DR: This paper analyzed DetectGPT and showed that selectively masking a combination of nouns, verbs, and adjectives improves the AUROC metric by up to 9.5%, demonstrating the importance of targeted masking strategies.