CS 153 · Stanford · Spring 2026

Discover what papers
actually disagree about

Upload scientific PDFs and ask a question. Surface real contradictions, hidden support, and apparent conflicts caused by terminology drift — not synthesized away.

3

Pipeline stages

F1 = 0.35

Benchmark F1

$0.003

Avg. cost / analysis

The problem: Tools like Perplexity and NotebookLM synthesize papers into confident summaries. For cross-domain research, this hides exactly the disagreements you need to see — especially when two papers use the same word to mean completely different things. ClaimLens makes those disagreements first-class outputs.

How it works

A three-stage LLM pipeline runs on your papers

1

Extract Claims

Each paper is broken into atomic, falsifiable claims tagged with verbatim source passages, key terms, and domain signals.

2

Normalize Terms

A cross-paper glossary flags where the same word means different things across fields — the root cause of most apparent conflicts.

3

Detect Conflicts

Claim pairs are classified with a plain-language explanation and, for incommensurable pairs, a terminology note.

SUPPORTClaims reinforce each otherCONTRADICTGenuine empirical or logical conflictINCOMMENSURABLEApparent conflict due to terminology drift

Try it yourself

Upload 2–10 scientific PDFs and ask a question

Drop PDFs here or click to upload

2–10 papers · PDF only