A machine-screening index for preprints

One preprint, three separate signals.

Evidence, Trust, and Novelty are recorded independently for every screening run—alongside the provenance that decides whether each one can be used at all.

screening runs in a corpus of indexed bioRxiv and medRxiv preprints.

A first-pass reading aid—not peer review, scientific verification, or a publication recommendation.

Live pipeline

Loading current pipeline data…

Live
Indexed
Screening runs
Eligible queue
Complete runs
Evidence
letter · A–E
Machine-detected strength of methods, results, and claim support. A = strongest recorded support; E = weakest.
Trust markers
letter · A–E
A meets nearly all applicable trust checks · C some · E minimal — transparency markers, reproducibility, statistical rigor, citation health; fraud checks cap the grade
Novelty
number · 1–10
1 landmark · 5 important · 10 minimal — machine-generated; lower = more novel
Recent screening signals
View all
Signal Paper Source Run Generated
Loading…

How it works

Automated checks, model-reviewer roles, explicit gates, and visible provenance.

01
Layer 1

Deterministic checks

Paper-mill detection, statcheck p-value verification, GRIM tests, data availability, retraction cross-referencing — before any LLM runs.

02
Layer 2

Eleven model-reviewer roles

Methods, statistics, ethics, validity, domain, adversarial, and positive-evidence roles inspect routed manuscript context. Their outputs are correlated machine signals, not independent peer reviews.

03
Layer 3

Rubric-guided synthesis

Structured model outputs become separate Evidence, Trust, and Novelty screening axes. Deterministic safety gates can cap or withhold signals when critical checks fail.

04
Layer 4

Provenance and abstention

Full-text input, role participation, automated-check health, judge execution, and literature grounding determine whether each axis is available, limited, or withheld.