arxiv-sanity-lite

Tag arXiv papers of interest and receive personalized recommendations via Support Vector Machines (SVMs) on tf-idf features of abstracts.

Arxiv-Sanity-Lite is Andrej Karpathy's efficient, open-source rewrite of the original Arxiv Sanity: it tames the arXiv paper flood. The web-based tool periodically polls the arXiv API, utilizing machine learning (SVMs over tf-idf features) and NLP to analyze and organize papers from categories like cs.[CV|AI|CL|LG|NE]. Users tag papers to build a personal library, and the system delivers personalized recommendations, including regular email digests of new, relevant research. This lighter, more scalable version focuses on core value: intelligent paper discovery and filtering.

https://arxiv-sanity-lite.com/