Reproduce “Host-Aware Web Crawling for LLM Pretraining Data” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Malte Ostendorff
The authors propose allocating crawler budgets by sampling pages from each host and crawling more deeply on hosts whose samples score well. In simulations over already available pages, this policy collected up to 7.8× as many FineWeb-Edu pages scoring above 0.5 as uniform breadth at small budgets. These results measure high-quality pages per fetch, not live-crawl or trained-model performance.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.