CLJul 26, 2019

RoBERTa: A Robustly Optimized BERT Pretraining Approach

arXiv:1907.11692v129858 citations
Originality Synthesis-oriented
AI Analysis

This work addresses the challenge of optimizing language model pretraining for researchers and practitioners, showing that careful hyperparameter tuning and data scaling can lead to significant performance gains, though it is incremental on BERT.

The authors conducted a replication study of BERT pretraining, finding that BERT was undertrained and that their optimized model (RoBERTa) achieves state-of-the-art results on benchmarks like GLUE, RACE, and SQuAD.

Language model pretraining has led to significant performance gains but careful comparison between different approaches is challenging. Training is computationally expensive, often done on private datasets of different sizes, and, as we will show, hyperparameter choices have significant impact on the final results. We present a replication study of BERT pretraining (Devlin et al., 2019) that carefully measures the impact of many key hyperparameters and training data size. We find that BERT was significantly undertrained, and can match or exceed the performance of every model published after it. Our best model achieves state-of-the-art results on GLUE, RACE and SQuAD. These results highlight the importance of previously overlooked design choices, and raise questions about the source of recently reported improvements. We release our models and code.

Code Implementations67 repos

Data from Papers with Code (CC-BY-SA-4.0)

Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes