Open EvidenceKit Lite: A Local Tool for First-Pass Citation Checking

A developer built Open EvidenceKit Lite, a local tool that takes an English claim and up to five synthetic evidence records and prepares a consistent first-pass review. The tool checks references, applies supplied time and source constraints, and uses an open-weight natural language inference model to suggest how each piece of evidence relates to the claim. Its reports combine source IDs, model scores, rule checks and review reasons, though the developer stresses that humans must still inspect the evidence and no time savings have been measured. The model's scores are uncalibrated and the tool does not authenticate sources or reliably extract dates, with missing references halting inference and conflicting evidence routed to human review. Testing on a wider candidate set showed weaker results, including a false support judgment on a quoted payment claim, and the developer chose to retain these failures in the public record. The project is MIT licensed and released on GitHub, with model weights downloaded separately.