Grading you can audit.
Polaris reads your policy and syllabus, scores every answer criterion by criterion, and shows the rule behind each mark. The teacher decides. Nothing is published without them.
Figures are the team’s reported results from the Polaris educator study, run with a faculty collaborator.
Pick an answer. Watch the marking.
Question: why do we split data into training, validation and test sets? (10 marks). Evaluate, then act as the teacher: change a mark, review anything flagged, publish.
Scripted demo on example answers. The deployed system runs an LLM agent over your own policy.
Shortlisted at the QS Reimagine Education Awards.
Vector Polaris, AI Innovation in Education. Awards and Conference 2026, London, 6 to 8 December. Polaris was built with a professor and tested with 36 educators in 19 institutions.

Ask VECTOR, and a calmer grading flow.
Questions answered from your own institutional data.
Retrieval over your own records. Answers stay inside the asker’s role.
Submit, propose, moderate, publish.
Assignments and written exam answers go through the same auto-evaluator. Marks land in the gradebook only after the teacher publishes.
VectorLMS
Book a demo