A paper reading regimen
How to keep up with fast-paced research.
The number of monthly submissions to the popular arXiv preprint server has reached 30.000 in March, 2026. There is a need to filter through the noise of irrelevant research articles.
My primary source for discovering new papers (as a machine learning researcher) is an RSS reader that aggregates multiple research feeds. Each day, I review the main feed and save promising papers for later reading in a reference management software. I use Feedly because it integrates RSS feeds and email newsletters in a single interface. For reference management, I use Zotero because it is free, open source, and supported by an extensive plugin ecosystem (e.g., Notero, Beaver).
I organize sources into two categories:
Original Sources
Google Scholar author alerts, work from relevant researchers
Google Scholar related-articles alerts, papers connected to relevant researchers
Google Scholar keyword alerts, specific topics and methods
Scholar Inbox, personalized paper recommendations
Journal alerts, high-impact publications
Feedly AI feeds, filtered research streams
News & Commentary
Research newsletters (e.g., The Batch, Hugging Face Daily Papers, Import AI, AI News, AI Newsletter, Causal Python), summaries of recent developments
Research blogs (e.g., Google Research, DeepMind, Anthropic, OpenAI), new research and industry trends