PAPER DISCOVERY
Start with a better question.
Explore search terms, related ideas, and relevant papers. Keep the title, author, and original source close to every promising lead.
Explore sample papersA little curiosity. A clearer picture.
Bring your research question into focus. Explore a thoughtful approach to AI-assisted paper discovery, literature organization, and evidence synthesis.
Built around questions. Grounded in sources.
Good research keeps the connections.
Question to paperPaper to findingFinding to sourceFOLLOW YOUR CURIOSITY
CluePebble explores the space between finding a relevant paper and understanding what a body of research actually says.
PAPER DISCOVERY
Explore search terms, related ideas, and relevant papers. Keep the title, author, and original source close to every promising lead.
Explore sample papersLITERATURE ORGANIZATION
Build a focused reading collection. Bring the research question, useful notes, and bibliographic details into the same view.
Build a sample collectionEVIDENCE SYNTHESIS
Compare methods and findings alongside their limits. Separate what a paper reports from the questions it leaves open.
Read an example evidence briefA SMALL NOTEBOOK. A USEFUL START.
Try a reading workflow with three real papers about retrieval and language models. Search, save a paper, and trace an example brief back to its sources.
Curated demo · Searches these 3 papers only. The evidence brief is a prepared example, not a live AI response.
Enable JavaScript to explore the interactive sample below. The linked papers are also listed in the evidence brief.
Saved in this browser. A focused starting point for your next reading session.
PREPARED EXAMPLE / ALL 3 SOURCE PAPERS
These papers address different parts of the same research question. They are useful to read together, but they are not a single head-to-head comparison.
DPR studies learned dense representations for passage retrieval in open-domain question answering.
[1] Karpukhin et al., 2020RAG combines a retriever with a sequence-to-sequence model, reporting gains on the knowledge-intensive tasks it evaluates.
[2] Lewis et al., 2020Lost in the Middle reports sensitivity to the position of relevant information in the long contexts and tasks it tests.
[3] Liu et al., 2023How do retrieval quality and evidence placement interact in your own task? These three abstracts alone do not settle that question.
Scope: editorial notes based on the linked abstracts, checked 19 November 2026. This example always covers all three papers, independently of your saved collection.
Small, deliberate steps. A reading trail you can retrace.
THE CLUEPEBBLE POINT OF VIEW
A pebble is a small thing. Put it in the right place, and it becomes a marker for the next person. That is how we think about research evidence.
BEFORE YOU BEGIN
The notebook on this website is an interactive demonstration using three curated papers. Search filters that small collection; saving and citation export work in your browser. It does not search a live scholarly database or generate new AI responses.
Yes. Each sample paper links to its original arXiv record. You can save papers and export BibTeX entries for those arXiv records. Check the final published version and your required citation style before using them in formal work.
When browser storage is available, the sample paper IDs and your language preference stay on this browser. They are not synced across devices. You can remove papers or clear the collection. Private browsing or browser cleanup may remove the saved state.
Use the contact section to prepare an email about your subject, sources, and intended output. It opens a draft for you to review and send through your own email application. Specific integrations and service availability can be discussed for your use case.
BRING YOUR NEXT QUESTION
Tell us about your research area and the part of the literature workflow you would like to make clearer.
support@cluepebble.comA conversation starts with a little context.