Novelty Scan

Is your research idea actually new?

Paste your idea. Novelty Scan searches the published literature and shows you the closest work that already exists.

Check my idea free
  • Real, linked papers
  • A 1–10 score
  • Written reasoning
Novelty report

Conformal joint abstention for multilingual scientific claim verification

Novelty score
6/10
Meaningful synthesis
Overall finding

A meaningful synthesis or new comparison

10papers shown
Reasoned reviewexcerpt

The idea rates 6 because its strongest overlap is with [3], which already fuses retrieval compatibility and verifier feedback for calibrated risk-controlled refusal, while [1] covers multilingual calibration…

Relevant papers1 of 10
Calibrated Selective Fact-Checking via Evidence Chain EvaluationDekun Yang · 2026

Large language models (LLMs) can achieve strong fact-checking accuracy, yet forced binary decisions conceal a critical reliability problem: systems may issue confident verdicts even when supporting evidence is weak, sparse, or internally inconsistent. We…

Six months in, somebody hands you the paper.

  • It used different words

    Semantic search finds it even when the terminology is not yours.

  • It sat outside your field

    The whole indexed record is searched, not your reading list.

  • It appeared last week

    Preprints indexed today are in scope. Last month's search is not.

  • It was buried on page nine

    Hundreds of candidates, ranked down to the ten closest to your idea.

A chatbot checks the first few results. This checks a few hundred.

What gets searched

The open web, through a handful of queries.

An academic index, through keyword, title and semantic queries.

How far it reads

The first few results it decides to open.

Hundreds of candidate papers, ranked down to the ten closest.

What comes back

Prose whose depth varies with how you happened to ask.

The same 1–10 rubric and the same sections on every run.

What you can check

Whichever sources it chose to mention.

Every paper it weighed, listed, linked and dated.

One paragraph in. One evidenced answer out.

  1. 01

    Describe your idea

    One paragraph: the problem, the method, and the part you think is new.

  2. 02

    The literature is searched

    Keyword and semantic search across Semantic Scholar, ranked by similarity to your idea.

  3. 03

    Read the scored review

    A 1–10 score, the reasoning behind it, and every paper it was weighed against.

Assistance, not a verdict.

No index covers everything. Novelty Scan finds the work you missed and shows the evidence — reading the papers, and the judgement, stay yours.

Start a scan
Novelty Scan — Research novelty, explained