On the Role of Bidirectionality in Language Model Pre-Training

• URL: http://arxiv.org/abs/2205.11726v1
• Date: Tue, 24 May 2022 02:25:05 GMT
• Title: On the Role of Bidirectionality in Language Model Pre-Training
• Authors: Mikel Artetxe, Jingfei Du, Naman Goyal, Luke Zettlemoyer, Ves Stoyanov
• Abstract summary: We study the role of bidirectionality in next token prediction, text infilling, zero-shot priming and fine-tuning. We train models with up to 6.7B parameters, and find differences to remain consistent at scale.
• Score: 85.14614350372004
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Prior work on language model pre-training has explored different architectures and learning objectives, but differences in data, hyperparameters and evaluation make a principled comparison difficult. In this work, we focus on bidirectionality as a key factor that differentiates existing approaches, and present a comprehensive study of its role in next token prediction, text infilling, zero-shot priming and fine-tuning. We propose a new framework that generalizes prior approaches, including fully unidirectional models like GPT, fully bidirectional models like BERT, and hybrid models like CM3 and prefix LM. Our framework distinguishes between two notions of bidirectionality (bidirectional context and bidirectional attention) and allows us to control each of them separately. We find that the optimal configuration is largely application-dependent (e.g., bidirectional attention is beneficial for fine-tuning and infilling, but harmful for next token prediction and zero-shot priming). We train models with up to 6.7B parameters, and find differences to remain consistent at scale. While prior work on scaling has focused on left-to-right autoregressive models, our results suggest that this approach comes with some trade-offs, and it might be worthwhile to develop very large bidirectional models.

Related papers

• Bidirectional Awareness Induction in Autoregressive Seq2Seq Models [47.82947878753809]
  Bidirectional Awareness Induction (BAI) is a training method that leverages a subset of elements in the network, the Pivots, to perform bidirectional learning without breaking the autoregressive constraints. In particular, we observed an increase of up to 2.4 CIDEr in Image-Captioning, 4.96 BLEU in Neural Machine Translation, and 1.16 ROUGE in Text Summarization compared to the respective baselines.
  arXiv  Detail & Related papers  (2024-08-25T23:46:35Z)
• Co-guiding for Multi-intent Spoken Language Understanding [53.30511968323911]
  We propose a novel model termed Co-guiding Net, which implements a two-stage framework achieving the mutual guidances between the two tasks. For the first stage, we propose single-task supervised contrastive learning, and for the second stage, we propose co-guiding supervised contrastive learning. Experiment results on multi-intent SLU show that our model outperforms existing models by a large margin.
  arXiv  Detail & Related papers  (2023-11-22T08:06:22Z)
• Improve Transformer Pre-Training with Decoupled Directional Relative Position Encoding and Representation Differentiations [23.2969212998404]
  We revisit the Transformer-based pre-trained language models and identify two problems that may limit the expressiveness of the model. Existing relative position encoding models confuse two heterogeneous information: relative distance and direction. We propose two novel techniques to improve pre-trained language models.
  arXiv  Detail & Related papers  (2022-10-09T12:35:04Z)
• Unifying Language Learning Paradigms [96.35981503087567]
  We present a unified framework for pre-training models that are universally effective across datasets and setups. We show how different pre-training objectives can be cast as one another and how interpolating between different objectives can be effective. Our model also achieve strong results at in-context learning, outperforming 175B GPT-3 on zero-shot SuperGLUE and tripling the performance of T5-XXL on one-shot summarization.
  arXiv  Detail & Related papers  (2022-05-10T19:32:20Z)
• DSEE: Dually Sparsity-embedded Efficient Tuning of Pre-trained Language Models [152.29364079385635]
  As pre-trained models grow bigger, the fine-tuning process can be time-consuming and computationally expensive. We propose a framework for resource- and parameter-efficient fine-tuning by leveraging the sparsity prior in both weight updates and the final model weights. Our proposed framework, dubbed Dually Sparsity-Embedded Efficient Tuning (DSEE), aims to achieve two key objectives: (i) parameter efficient fine-tuning and (ii) resource-efficient inference.
  arXiv  Detail & Related papers  (2021-10-30T03:29:47Z)
• Unsupervised Paraphrasing with Pretrained Language Models [85.03373221588707]
  We propose a training pipeline that enables pre-trained language models to generate high-quality paraphrases in an unsupervised setting. Our recipe consists of task-adaptation, self-supervision, and a novel decoding algorithm named Dynamic Blocking. We show with automatic and human evaluations that our approach achieves state-of-the-art performance on both the Quora Question Pair and the ParaNMT datasets.
  arXiv  Detail & Related papers  (2020-10-24T11:55:28Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.