Seon Min Kim
TeamCollaboration

AuData

A research-integrity auditor that checks a single paper for statistical errors, numerical inconsistencies, figure manipulation and citation problems, with a human reviewing every flag.

Context
Collaborative repository (haile-teshome/AuData)
Team
Team
Tools
LLM agents, PubMed, Crossref, OpenAlex, Next.js
Links
GitHub

What it does

AuData audits one paper or preprint and surfaces prioritized, evidence-linked flags for statistical recomputation, internal number consistency, image forensics, methods versus claims, and reference integrity. It is built as a reviewer aid: every flag goes to a human, and nothing is an automated accusation.

Pipeline

Manage → Ingest → Detect → Reliability → Report. The reliability layer calibrates each flag and lets the system abstain instead of guessing.