Paperscope

Paperscope indexes around 6,000 AI and machine learning papers from arXiv, enriched with citation counts from Semantic Scholar. Type a title, an author, or describe an idea in your own words: Meilisearch matches both the words and the meaning. Open any paper to find related work, then ask questions across the whole corpus, your reading list, or a single paper.

​

Key features

• Hybrid search slider: Move from exact words to meaning in real time by changing semanticRatio. The results header shows how many papers were found only by meaning, using semanticHitCount.
• Built-in embeddings: Documents and queries are embedded inside Meilisearch with a HuggingFace model (BAAI/bge-small-en-v1.5), with no external embedding service.
• Similar papers: Each paper lists its nearest neighbours from the /similar endpoint, optionally filtered to the same arXiv category.
• Faceted navigation: Filter by category, topic, author, and year. The year slider uses facetStats, and the author box uses facet search.
• Citation-aware ranking: A custom ranking rule on citationCount breaks ties in favor of influential papers, and results can be sorted by citations or date.
• Multi-search: Papers and matching authors come back in a single request across two indexes.
• Chat with the papers: The “Ask the papers” page uses the /chats route. The LLM runs hybrid searches as a tool and streams its progress and sources.
• Scoped chat with tenant tokens: Chatting with a reading list or a single paper uses a short-lived tenant token whose search rule restricts retrieval to those papers.

​

Links