Open AccessJournal Article
Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer
Colin Raffel,Noam Shazeer,Adam Roberts,Katherine Lee,Sharan Narang,Michael Matena,Yanqi Zhou,Wei Li,Peter J. Liu +8 more
Reads0
Chats0
TLDR
This article introduced a unified framework that converts all text-based language problems into a text-to-text format and compared pre-training objectives, architectures, unlabeled data sets, transfer approaches, and other factors on dozens of language understanding tasks.Abstract:
Transfer learning, where a model is first pre-trained on a data-rich task before being fine-tuned on a downstream task, has emerged as a powerful technique in natural language processing (NLP). The effectiveness of transfer learning has given rise to a diversity of approaches, methodology, and practice. In this paper, we explore the landscape of transfer learning techniques for NLP by introducing a unified framework that converts all text-based language problems into a text-to-text format. Our systematic study compares pre-training objectives, architectures, unlabeled data sets, transfer approaches, and other factors on dozens of language understanding tasks. By combining the insights from our exploration with scale and our new ``Colossal Clean Crawled Corpus'', we achieve state-of-the-art results on many benchmarks covering summarization, question answering, text classification, and more. To facilitate future work on transfer learning for NLP, we release our data set, pre-trained models, and code.read more
Citations
More filters
Proceedings Article
Joint Passage Ranking for Diverse Multi-Answer Retrieval
TL;DR: This article propose a joint passage retrieval model focusing on re-ranking, which makes use of an autoregressive re-ranker that selects a sequence of passages, equipped with novel training and decoding algorithms.
Posted Content
Can Vision Transformers Perform Convolution
TL;DR: In this paper, a self-attention layer of a single ViT layer with image patches as the input can perform any convolution operation constructively, where the multi-head attention mechanism and the relative positional encoding play essential roles.
Proceedings Article
Bandits Don’t Follow Rules: Balancing Multi-Facet Machine Translation with Multi-Armed Bandits
TL;DR: The authors propose a multi-armed bandit to dynamically choose between different facets in a way that is most beneficial for the MT system and evaluate it on three different multi-facet applications: balancing translationese and natural training data.
Proceedings Article
Comprehensive Punctuation Restoration for English and Polish.
Michal Pogoda,Tomasz Walkowiak +1 more
TL;DR: The authors evaluated several methods in the comprehensive punctuation reconstruction task in the parallel corpora of two different languages, English and Polish, with a relatively simple and complex morphology, respectively, and also investigated the influence of building a model on the quality of the basic punctuation restoration task.
Posted Content
Wacky Weights in Learned Sparse Representations and the Revenge of Score-at-a-Time Query Evaluation
TL;DR: In this paper, the authors compare the performance of score-at-a-time query evaluation techniques for the top-k retrieval problem, and find that the former approach yields lower mean query latency, whereas the latter approach has more predictable query latency.