Search and Filter

Simulating LLM-to-LLM Tutoring for Multilingual Math Feedback

Date

AI Overview

Why it matters

AI models can generate instructional hints in English, but their ability to provide effective math feedback across diverse languages remains largely unexamined, risking educational inequities for non-English learners.

Study at a glance
  • Question: Does multilingual feedback from AI tutors lead to measurable learning gains in student language models?
  • Setting and data: 250 math word problems translated into 11 languages from the Multilingual Grade School Math benchmark.
  • Method: Model or benchmark evaluation simulating student-tutor interactions across 352 experimental configurations.
  • AI tools: LLaMA-3.3-70B, LLaMA-3.1-8B, Mistral-7B, Aya-8B, GPT-4o, Deepseek-R1-LLaMA-distill-70B
Key findings
  • Hints generated in English consistently produced the highest student gains across both high- and low-resource languages, with an overall average gain of 23.90% under English-only prompting.
  • The larger tutor model, LLaMA-3.3-70B, consistently outperformed the smaller LLaMA-3.1-8B model, especially for low-resource languages where median student gains reached up to 38%.
  • Multilingual student models gained more from hints on low-resource languages, whereas monolingual student models performed better on high-resource languages.

Summary drafted with AI from the full paper and may contain errors. See the full paper for details.

Large language models (LLMs) have demonstrated the ability to generate formative feedback and instructional hints in English, making them increasingly relevant for AI-assisted education. However, their ability to provide effective instructional support across different languages, especially for mathematically grounded reasoning tasks, remains largely unexamined. In this work, we present the first large-scale simulation of multilingual tutor-student interactions using LLMs. A stronger model plays the role of the tutor, generating feedback in the form of hints, while a weaker model simulates the student. We explore 352 experimental settings across 11 typologically diverse languages, four state-of-the-art LLMs, and multiple prompting strategies to assess whether language-specific feedback leads to measurable learning gains. Our study examines how student input language, teacher feedback language, model choice, and language resource level jointly influence performance. Results show that multilingual hints can significantly improve learning outcomes, particularly in low-resource languages when feedback is aligned with the student's native language. These findings offer practical insights for developing multilingual, LLM-based educational tools that are both effective and inclusive.