Multi-agent teacher-in-the-loop architecture for context personalization

Design · All designs
Evidence · no claims cited

Description

The article introduces a multi-agent, teacher-in-the-loop architecture for context personalization in which a teacher enters a base problem and desired topic, an LLM generates the problem, and then four AI agents evaluate it using criteria each specializes in: mathematical accuracy, authenticity, readability, and realism. Each agent has a specified role and prompt and returns a binary pass/fail decision. Failed cases trigger a refinement agent that revises the problem, iterating until all evaluations pass or a maximum of five refinement steps is reached. After refinement, teachers can further author the problem before submitting it.

Design Implications

Context

Requirements

• A teacher provides a base problem and a topic, and can choose whether to retain or modify the original numerical values

Constraints

• The process iterates until all agent evaluations pass or a maximum number of refinement steps is reached (five in this study)

Target Learners

• Middle school mathematics students (7th grade)

Learning Goals

• Personalized mathematics word problems aligned to student interests

Claims

Related Designs

• Four specialized evaluation agents for LLM-generated problems
• Multi-agent generate–validate–revise framework with four specialized validator agents for math problem personalization
• MAIC: an LLM-driven multi-agent online course framework balancing scalability and adaptivity

Examples

-

Key Sources

• Walkington, C., Beauchamp, T., Ikram, F., Koçyiğit Gürbüz, M., Xia, F., Lee, M., Lan, A. (2025). Mathematics Teachers’ Interactions with a Multi-Agent System for Personalized Problem Generation. doi:10.1007/978-3-032-29770-9_5

Page metadata

Field	Value
type	design
id	multi-agent-teacher-in-the-loop-architecture
title	Multi-agent teacher-in-the-loop architecture for context personalization
description	The article introduces a multi-agent, teacher-in-the-loop architecture for context personalization in which a teacher enters a base problem and desired topic, an LLM generates the problem, and then four AI agents eval...
status	review
generated	by process:settle-candidates on 2026-10-12
trust	unverified — no one has confirmed this page's content
sources	1 — walkington-2025

Frontmatter is the page's machine-readable record — see the Schema & Guide.