We automated the screening process for a literature review https://t.co/5jTufZPABX. With lumina's API, you can get structured output over millions of papers based on your search criteria. Researchers spend weeks doing reviews like this, we help them bring it down to minutes.

More results here:

Lumina serves queries from around the world, in the user's native language

Access to science is bottlenecked by language. Over 75% of people on the internet are not English speakers. We compared Google Scholar and Semantic Scholar to Lumina for French, Arabic, Chinese and Spanish questions, and saw that we were 6x more relevant on benchmarks.

More results here:

Lumina serves users from across the world, and lets users search for publications in their OWN language.

i love

thank you!

Here's a quick demo!

Our database covers a wide range of topics and fields, all pre-labeled in our dataset. You can search for papers across any domain or repository like PubMed and arXiv, using our powerful filters.

https://t.co/OpBjU26F8C is 5x more relevant than Google Scholar. To achieve this, we’ve grown our database from 3.5M to 100M embedded research objects to create the most performant search over scientific literature. Here's a breakdown of what we've indexed 👇

Now that we have the best search for scientific literature, we plan on bringing our SOTA infra to more datasets. Much much more to come! For more updates, or to leave feedback, please join our discord!

Not to mention all of the other awesome tooling we used to get this done! Thank you to @modal_labs @qdrant_engine.

We wouldn’t have gotten this done without help from our friends and providers. We want to give a huge shout out to @trieveai @skeptrune @cdxker @_federicoct @DSumesh4661 for spending countless all nighters with us to build this.

With GPT-4o as a judge, we compared the relevance of each search result for lumina, google scholar and semantic scholar. We’ve open sourced our benchmark here: https://t.co/crcbNbn2wD We hope to improve this over the next few months.

We will be releasing an API, that gives developers (and AI), the ability to build with our data to automate tedious research tasks. To get early access to the API, feel free to reach out here:

Now that LLMs can browse and research autonomously, it's even more important to have reliable, fast and highly accurate endpoints to interface with. With the Lumina API and a simple recursion script, you can get over 11x relevance over Google Scholar.

Google scholar only returns 1 highly relevant result, for only 50% of queries! There’s no point in having data if you can’t search it properly. We believe this is the case for most search tooling, and that there is room to make search better across the board.

We’ve brought SOTA search techniques to the most dense and valuable set of human knowledge. With lumina, you will find research that you wouldn’t have found otherwise. We search over >500M vectors, 1T embedded words, in less than 500ms.

Today we're announcing https://t.co/OpBjU26F8C, a SOTA search engine for scientific literature. Lumina searches over 100M research objects and delivers search results that are 5x more relevant than Google Scholar’s on average, without the aid of LLMs. It's completely free.