Searchthearxiv

Search the arXiv is a semantic search engine for machine learning papers on arXiv. Instead of matching keywords, it lets a researcher describe what they are looking for in plain language, or paste an arXiv link to find papers similar to one they already know, and returns the closest matches. The site embeds the abstracts of more than 500,000 papers using OpenAI's ada 002 text embedding model. The papers are drawn from the arXiv metadata set on Kaggle and limited to the categories cs.CV, cs.LG, cs.CL, cs.AI, cs.NE and cs.RO. When a search runs, the query is embedded with the same model and the 100 papers with the highest cosine similarity are returned. Results can be sorted by similarity, citation count, newest or oldest, and the interface also offers a people view alongside papers. Citation counts come from the Semantic Scholar API. It suits students, engineers and researchers who need to survey a topic, find related work for a paper they are reading, or track down a half remembered idea without knowing the exact terms. Coverage is limited to the machine learning and related categories listed above, so it is not a general index of all arXiv subjects. The site describes itself as a personal project and an experiment. The source code is on GitHub and the embeddings are published on Kaggle. No pricing is published and the site has no paid plans; it carries a donate link instead.