AI Paper Similarity Search

How to Use the AI Paper Similarity Search

• Enter a query (e.g., "computer vision") in the search box.
• Select the number of paper embeddings to load from the dropdown menu.
• Click the "Find Similar Papers" button to search for papers that match your query.
• The most similar papers will be displayed and ranked with the title, similarity score, and a link to the paper on arXiv.
• You can click "Read More" to expand the abstract of each paper.

This is a client-side application powered by ONNX Runtime, Transformers.js, and arXiv dataset.

Note: Only a subset of the dataset is loaded due to the size of the embeddings (20+ GB). To access the entire collection of 2.6 million paper embeddings, please visit the GitHub repository and run the model on a server.