Challenge: a dominant practice is to fine tune large pretrained transformer models using smaller downstream datasets . performance gains are not always attributable to the use of external data in massive amounts .
Approach: They propose to use the same (downstream) training data for pretraining and finetuning to compare models.
Outcome: The proposed model outperforms standard pretraining on the BookWiki corpus on 7 and 5 datasets.

Similar Papers

How much pretraining data do language models need to learn syntax? (2021.emnlp-main)

Copied to clipboard

Challenge: Pretraining methods are convenient, but expensive in terms of time and resources.
Approach: They investigate the impact of pretraining data size on the syntactic capabilities of RoBERTa by using syntaktic structural probes to determine whether models pretrained on more data encode a higher amount of syntastic information.
Outcome: The proposed models perform better on part-of-speech tagging, dependency parsing and paraphrase identification.
Does Pretraining for Summarization Require Knowledge Transfer? (2021.findings-emnlp)

Copied to clipboard

Challenge: Existing theories claim that pretraining models learn linguistic knowledge from the pretraining corpus, but scientific explanations for these benefits remain unknown.
Approach: They propose to use random character n-grams to test models on real corpora to see if the small residual benefit of using real data could be accounted for by the structure of the pretraining task.
Outcome: The proposed task performs on documents consisting of character n-grams, whereas pretrained models perform on real corpora with no residual benefit.
How Much Pretraining Does Structured Data Need? (2026.eacl-long)

Copied to clipboard

Challenge: Large language models are increasingly adopted for handling structured data, despite pretraining on unstructured text.
Approach: They propose to re-initialize subsets of layers with random weights before fine-tuning on structured datasets.
Outcome: The proposed models are compared to unstructured datasets and show that they perform well over structured data.
To Pretrain or Not to Pretrain: Examining the Benefits of Pretrainng on Resource Rich Tasks (2020.acl-main)

Copied to clipboard

Challenge: Existing studies on pretraining NLP models with variants of Masked Language Model (MLM) objectives have shown that the number of training samples used in the downstream task is limited.
Approach: They propose to use MLM objectives to pretrain NLP models with variants of Masked Language Model (MLM) objectives to improve accuracy on downstream tasks.
Outcome: The proposed model can reach a diminishing return point as the supervised data size increases significantly.
Data Contamination: From Memorization to Exploitation (2022.acl-short)

Copied to clipboard

Challenge: Pretrained language models are typically trained on web-based datasets that are often "contaminated" with downstream test sets.
Approach: They propose a method to pretrain BERT models on Wikipedia and labeled downstream datasets and fine-tune them on the relevant task.
Outcome: The proposed method compares models on Wikipedia and labeled downstream datasets on two models and three downstream tasks.
Cost-effective Selection of Pretraining Data: A Case Study of Pretraining BERT on Social Media (2020.findings-emnlp)

Copied to clipboard

Challenge: Recent studies show that domain-specific BERT models can be improved when in-domain data is used for pretraining.
Approach: They propose to use Twitter and forum text as pretraining sources for two BERT models and use similarity measures to nominate in-domain data for pretraining.
Outcome: The proposed method can be used to improve performance on downstream tasks by using in-domain data.
Infusing Finetuning with Semantic Dependencies (2021.tacl-1)

Copied to clipboard

Challenge: Several diagnostics help to localize the benefits of our approach.
Approach: They apply convolutional graph encoders to integrate semantic parses into task-specific finetuning.
Outcome: The proposed approach yields benefits to natural language understanding (NLU) tasks in the GLUE benchmark.
Go Simple and Pre-Train on Domain-Specific Corpora: On the Role of Training Data for Text Classification (2020.coling-main)

Copied to clipboard

Challenge: Pre-trained language models provide the foundations for state-of-the-art performance across a wide range of natural language processing tasks, including text classification.
Approach: They compare the performance of a linear classifier based on word embeddings with a pre-trained language model, i.e., BERT, across a wide range of datasets and classification tasks.
Outcome: The proposed method outperforms baselines in standard datasets with large training sets, but in settings with small training datasets it performs better.
Deciphering the Impact of Pretraining Data on Large Language Models through Machine Unlearning (2024.findings-acl)

Copied to clipboard

Challenge: Existing studies have suggested that the composition of the pretraining corpus exerts a significant impact upon the performance of LLMs.
Approach: They analyze the impact of 48 datasets from 5 major categories of pretraining data of Large Language Models and measure their impacts on LLMs using benchmarks about nine major categories.
Outcome: The proposed analysis provides insights into the organization of data to support more efficient pretraining of Large Language Models.
Data, Data Everywhere: A Guide for Pretraining Dataset Construction (2024.emnlp-main)

Copied to clipboard

Challenge: Recent language models have impressive capabilities on a number of evaluation areas.
Approach: They conduct systematic analysis of pretraining set construction to identify which methods yield the greatest gains in model accuracy.
Outcome: The proposed method can be used to refine and improve a pretraining set.

What is GenGO?

GenGO is an NLP powered publication search system. It currenctly indexes 30k+ papers from ACL Anthology, and implements multi-aspect summarization, semantic search, and more!

Information

About
Limitations