Glite Get Started
Glite menu
Research

Anomalous Space, 38 Pentonville Road, London N1 9HF18:30 BST

Glite’s learning architecture and the models behind concept extraction.

The audience and opening remarks in the meetup room at Anomalous Space on 6 October 2026.
The audience and opening remarks at Anomalous Space, 6 October 2026. Video still.
First pages of papers discussed at the October 2026 meetup.
Papers discussed at the October meetup.

Nine papers on AI in education

Recent studies of classroom AI, student simulation, knowledge tracing and adaptive teaching.

Glite’s original N+1 learning-system diagram, connecting the student, learning history, concept extractor, knowledge predictor and orchestrator.
The N+1 learning-system architecture, from Pavel’s presentation.

The N+1 approach to education

Pavel Katunin · Glite

Identifying concepts, predicting a learner’s knowledge and choosing the next useful task.

The hybrid concept-extraction model first retrieves candidate concepts using embeddings, then ranks them with a scoring model.
The hybrid model, from Vassili’s presentation.

How do we build a concept extractor?

Vassili Philippov · Glite

The models, labels and costs behind identifying learning concepts. The article includes the opening quiz.

Paper on arXiv