CoLearn deployable containerised web system¶
Product or Programme · All products and programmes
Evidence · 2 claims (2 for) · 1 study (1 design),q2· 0 of 1 report an effect size · 2 claims rest on one study
Description¶
CoLearn is a containerised web application developed by Kailai He et al. for adaptive diagnostic Biology and Chemistry practice with LLM-generated questions and misconception feedback.
Components¶
- CoLearn deployable containerised web system (Vite/React SPA, FastAPI backend, provider-selectable LLM): CoLearn is a container-deployed web application whose frontend is a Vite + React + TypeScript single-page app and whose backend is FastAPI with an async MongoDB store, JWT/bcrypt authentication, and SSE streaming for live generation and grading. The LLM backend is provider-selectable (AWS Bedrock with Claude Sonnet 4.5 as the live default; an OpenAI-compatible path is also supported). "The whole stack starts with one command via Docker Compose", and the backend ships with a test suite covering BKT, grading, generation, streaming, RBAC, and the A/B analytics. (Kailai He et al. (2026))
- Make an adaptive tutor's personalisation visible and testable through live mastery visualisation and built-in blind A/B comparison: CoLearn renders the agent's evolving belief in real time so "personalisation is something the user can see rather than take on trust", and includes a blind A/B interface presenting personalised and non-personalised questions in randomised, unlabelled positions. Results aggregate into an analytics dashboard reporting the preference rate, a 95% confidence interval, and a two-sided sign test against the 0.5 chance rate, making the adaptation inspectable and empirically testable by anyone. (Kailai He et al. (2026))
Claims¶
- In blind A/B comparison, raters prefer personalised (Adaptive) questions over non-personalised (Random-topic) questions about 68–69% of the time in Biology and Chemistry [+M]
- In a 16-rater usability survey after a session, CoLearn scores an overall mean of 4.0/5 with no dimension below ≈3.7, and 75% (12/16) would reuse it; Responsiveness is the lowest dimension [+W]
Related Products and Programmes¶
Key Sources¶
- Kailai He, Zhihao Wu, Linhai Zhang, Runcong Zhao, Yulan He, Jiazheng Li. (2026). CoLearn: An Agentic Tutor that Learns its Learner in a Human–AI Co-Learning Loop. https://arxiv.org/abs/2609.21154
Page metadata
| Field | Value |
|---|---|
type |
product |
id |
colearn-deployable-containerised-web-system |
title |
CoLearn deployable containerised web system |
description |
CoLearn is a containerised web application developed by Kailai He et al. |
status |
review |
product_kind |
software |
generated |
by process:settle-candidates on 2026-10-10 |
trust |
unverified — no one has confirmed this page's content |
Frontmatter is the page's machine-readable record — see the Schema & Guide.