Lens

Who Lens is for

For researchers holding forty papers on one topic who need to know which of them said the thing they half-remember. Cross-document question answering is the feature that turns a folder of PDFs into something you can interrogate.

What Lens does

Upload research PDFs and Lens extracts and embeds the text, then lets you read each one in the browser and save annotations against specific passages. Questions can be scoped to a single paper or asked across the whole collection, with answers grounded in retrieved passages rather than in the model’s general knowledge — so the answer points at where it came from and you can check it. That checkability is the entire point when the material is research you intend to cite.

• →In-browser PDF reading with passage annotation
• →Single-document and cross-document question answering
• →Answers grounded in retrieved passages you can verify
• →Text extracted and embedded on upload

How Lens works

1. 01

   Upload your PDFs

   Text is extracted and embedded so passages can be retrieved by meaning rather than by keyword.

2. 02

   Read and annotate

   Read in the browser and save annotations against specific excerpts, so a passage and your reaction to it stay together.

3. 03

   Ask one paper a question

   Scoped to a single document when you know which one you need.

4. 04

   Ask the whole collection

   Cross-document questions answered from retrieved passages, which is how you find the paper you half-remember.

What Lens does not do

Lens reads PDFs with an extractable text layer. A scanned image with no text layer will not work, because there is no OCR — a real limit for older papers. It does not fetch papers for you and has no publisher integration, so you supply the files. Answers are grounded in retrieved passages, which means a passage that retrieval misses is a passage the answer will not reflect; verify before citing.

Common questions

Works well with