Is it really in the paper?
Saw a big claim about an AI paper on LinkedIn or in the news? Paste the post and the paper’s link. Every claim gets checked against the paper’s own abstract, with the exact line as proof.
The Examiner opens soon. Until then, every line in the morning digest is checked the same way.
- Fetching the abstract from arXiv
- Pulling the claims out of the post
- Checking each claim against the abstract
- Making sure every quote is really there
How the Examiner decides
- In the paper
- The abstract clearly says it.
- Stretched
- The abstract says something close, but the claim goes further: bigger numbers, stronger words, or a wider reach.
- Paper says otherwise
- The abstract says the opposite.
- Not in the abstract
- The abstract doesn’t say it. That doesn’t make it false, but it needs a better source than this paper’s summary.
An AI model suggests each verdict and copies the line that proves it. Then plain code checks that the line really is in the abstract (word for word, numbers included) before anything is shown. A made-up quote can’t pass, so a claim without real proof never shows as proven. It’s the same check that guards the morning digest.