Synthetic continued pretraining

• URL: http://arxiv.org/abs/2409.07431v2
• Date: Thu, 3 Oct 2024 13:07:25 GMT
• Title: Synthetic continued pretraining
• Authors: Zitong Yang, Neil Band, Shuangping Li, Emmanuel Candès, Tatsunori Hashimoto,
• Abstract summary: We propose synthetic continued pretraining on a small corpus of domain-specific documents. We instantiate this proposal with EntiGraph, a synthetic data augmentation algorithm. We show how synthetic data augmentation can "rearrange" knowledge to enable more data-efficient learning.
• Score: 29.6872772403251
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Pretraining on large-scale, unstructured internet text enables language models to acquire a significant amount of world knowledge. However, this knowledge acquisition is data-inefficient--to learn a given fact, models must be trained on hundreds to thousands of diverse representations of it. This poses a challenge when adapting a pretrained model to a small corpus of domain-specific documents, where each fact may appear rarely or only once. We propose to bridge this gap with synthetic continued pretraining: using the small domain-specific corpus to synthesize a large corpus more amenable to learning, and then performing continued pretraining on the synthesized corpus. We instantiate this proposal with EntiGraph, a synthetic data augmentation algorithm that extracts salient entities from the source documents and then generates diverse text by drawing connections between the sampled entities. Synthetic continued pretraining with EntiGraph enables a language model to answer questions and follow generic instructions related to the source documents without access to them. If, instead, the source documents are available at inference time, we show that the knowledge acquired through our approach compounds with retrieval-augmented generation. To better understand these results, we build a simple mathematical model of EntiGraph, and show how synthetic data augmentation can "rearrange" knowledge to enable more data-efficient learning.

Related papers

• Fake it to make it: Using synthetic data to remedy the data shortage in joint multimodal speech-and-gesture synthesis [21.210982054134686]
  Methods for joint and unified synthesis of speech audio and co-speech 3D gesture motion from text are a new and emerging field. Existing methods are trained on parallel data from all constituent modalities. Inspired by student-teacher methods, we propose a straightforward solution to the data shortage, by simply synthesising additional training material.
  arXiv  Detail & Related papers  (2024-04-30T15:22:19Z)
• Dolma: an Open Corpus of Three Trillion Tokens for Language Model Pretraining Research [139.69207791947738]
  Dolma is a three-trillion-token English corpus built from a diverse mixture of web content, scientific papers, code, public-domain books, social media, and encyclopedic materials. We document Dolma, including its design principles, details about its construction, and a summary of its contents. We present analyses and experimental results on intermediate states of Dolma to share what we have learned about important data curation practices.
  arXiv  Detail & Related papers  (2024-01-31T20:29:50Z)
• Image Captions are Natural Prompts for Text-to-Image Models [70.30915140413383]
  We analyze the relationship between the training effect of synthetic data and the synthetic data distribution induced by prompts. We propose a simple yet effective method that prompts text-to-image generative models to synthesize more informative and diverse training data. Our method significantly improves the performance of models trained on synthetic training data.
  arXiv  Detail & Related papers  (2023-07-17T14:38:11Z)
• Synthetic Pre-Training Tasks for Neural Machine Translation [16.6378815054841]
  Our goal is to understand the factors that contribute to the effectiveness of pre-training models when using synthetic resources. We propose several novel approaches to pre-training translation models that involve different levels of lexical and structural knowledge. Our experiments on multiple language pairs reveal that pre-training benefits can be realized even with high levels of obfuscation or purely synthetic parallel data.
  arXiv  Detail & Related papers  (2022-12-19T21:34:00Z)
• Schema-aware Reference as Prompt Improves Data-Efficient Knowledge Graph Construction [57.854498238624366]
  We propose a retrieval-augmented approach, which retrieves schema-aware Reference As Prompt (RAP) for data-efficient knowledge graph construction. RAP can dynamically leverage schema and knowledge inherited from human-annotated and weak-supervised data as a prompt for each sample.
  arXiv  Detail & Related papers  (2022-10-19T16:40:28Z)
• KELM: Knowledge Enhanced Pre-Trained Language Representations with Message Passing on Hierarchical Relational Graphs [26.557447199727758]
  We propose a novel knowledge-aware language model framework based on fine-tuning process. Our model can efficiently incorporate world knowledge from KGs into existing language models such as BERT.
  arXiv  Detail & Related papers  (2021-09-09T12:39:17Z)
• Integrating Semantics and Neighborhood Information with Graph-Driven Generative Models for Document Retrieval [51.823187647843945]
  In this paper, we encode the neighborhood information with a graph-induced Gaussian distribution, and propose to integrate the two types of information with a graph-driven generative model. Under the approximation, we prove that the training objective can be decomposed into terms involving only singleton or pairwise documents, enabling the model to be trained as efficiently as uncorrelated ones.
  arXiv  Detail & Related papers  (2021-05-27T11:29:03Z)
• Improving Text Relationship Modeling with Artificial Data [0.07614628596146598]
  We apply and evaluate a synthetic data approach to relationship classification in digital libraries. We find that for classification on whole-part relationships between books, synthetic data improves a deep neural network classifier by 91%.
  arXiv  Detail & Related papers  (2020-10-27T22:04:54Z)
• Leveraging Graph to Improve Abstractive Multi-Document Summarization [50.62418656177642]
  We develop a neural abstractive multi-document summarization (MDS) model which can leverage well-known graph representations of documents. Our model utilizes graphs to encode documents in order to capture cross-document relations, which is crucial to summarizing long documents. Our model can also take advantage of graphs to guide the summary generation process, which is beneficial for generating coherent and concise summaries.
  arXiv  Detail & Related papers  (2020-05-20T13:39:47Z)
• Knowledge Graph-Augmented Abstractive Summarization with Semantic-Driven Cloze Reward [42.925345819778656]
  We present ASGARD, a novel framework for Abstractive Summarization with Graph-Augmentation and semantic-driven RewarD. We propose the use of dual encoders---a sequential document encoder and a graph-structured encoder---to maintain the global context and local characteristics of entities. Results show that our models produce significantly higher ROUGE scores than a variant without knowledge graph as input on both New York Times and CNN/Daily Mail datasets.
  arXiv  Detail & Related papers  (2020-05-03T18:23:06Z)
• Exploiting Structured Knowledge in Text via Graph-Guided Representation Learning [73.0598186896953]
  We present two self-supervised tasks learning over raw text with the guidance from knowledge graphs. Building upon entity-level masked language models, our first contribution is an entity masking scheme. In contrast to existing paradigms, our approach uses knowledge graphs implicitly, only during pre-training.
  arXiv  Detail & Related papers  (2020-04-29T14:22:42Z)

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.