Craw4LLM: Efficient Web Crawling for LLM Pretraining
This work addresses the problem of high waste and low data quality in web crawling for LLM pretraining, offering a more efficient approach for researchers and developers in AI and NLP.
The paper tackles the inefficiency of web crawling for LLM pretraining by introducing Craw4LLM, a method that prioritizes webpages based on their influence on LLM performance rather than standard graph connectivity, achieving the same downstream performance with only 21% of URLs crawled.
Web crawl is a main source of large language models' (LLMs) pretraining data, but the majority of crawled web pages are discarded in pretraining due to low data quality. This paper presents Craw4LLM, an efficient web crawling method that explores the web graph based on the preference of LLM pretraining. Specifically, it leverages the influence of a webpage in LLM pretraining as the priority score of the web crawler's scheduler, replacing the standard graph connectivity based priority. Our experiments on a web graph containing 900 million webpages from a commercial search engine's index demonstrate the efficiency of Craw4LLM in obtaining high-quality pretraining data. With just 21% URLs crawled, LLMs pretrained on Craw4LLM data reach the same downstream performances of previous crawls, significantly reducing the crawling waste and alleviating the burdens on websites. Our code is publicly available at https://github.com/cxcscmu/Craw4LLM.