Adapting Large Language Models for Education: Foundational Capabilities, Potentials, and Challenges

TL;DR

<3-5 sentence high-level summary> This survey addresses the problem of delivering personalized, dialog-enabled education at scale by analyzing five foundational LLM-enabled capabilities—mathematics, writing, programming, reasoning, and knowledge-based question answering—and two architectural directions (unified LLM vs mixture-of-experts with an LLM controller). It surveys current progress, benchmarks, and methodological trends (data quality, prompting, tool integration, and retrieval) for building an intelligent education system. Key contributions include capability-centered analysis, comparison with prior surveys, and design considerations for unified versus MoE-based frameworks, along with discussions of practical challenges such as hallucination, bias, and multi-modal data. The findings highlight that while GPT-4 and other large models show strong overall performance, no single model excels across all capabilities, underscoring the need for hybrid architectures and targeted data to realize effective LLM-based education systems with real-world impact.

Abstract

Online education platforms, leveraging the internet to distribute education resources, seek to provide convenient education but often fall short in real-time communication with students. They often struggle to address the diverse obstacles students encounter throughout their learning journey. Solving the problems encountered by students poses a significant challenge for traditional deep learning models, as it requires not only a broad spectrum of subject knowledge but also the ability to understand what constitutes a student's individual difficulties. It's challenging for traditional machine learning models, as they lack the capacity to comprehend students' personalized needs. Recently, the emergence of large language models (LLMs) offers the possibility for resolving this issue by comprehending individual requests. Although LLMs have been successful in various fields, creating an LLM-based education system is still challenging for the wide range of educational skills required. This paper reviews the recently emerged LLM research related to educational capabilities, including mathematics, writing, programming, reasoning, and knowledge-based question answering, with the aim to explore their potential in constructing the next-generation intelligent education system. Specifically, for each capability, we focus on investigating two aspects. Firstly, we examine the current state of LLMs regarding this capability: how advanced they have become, whether they surpass human abilities, and what deficiencies might exist. Secondly, we evaluate whether the development methods for LLMs in this area are generalizable, that is, whether these methods can be applied to construct a comprehensive educational supermodel with strengths across various capabilities, rather than being effective in only a singular aspect.

Figures (4)

• Figure 1: An example of LLM-based education systems integrating multiple abilities to solve student problems.
• Figure 2: Summary of LLM's education-related foundational capabilities.
• Figure 3: A summary framework diagram for the approaches of LLMs in the development of education-related abilities. It categorizes previous enhancement strategies into three parts: Input Data Refinement, Model Self-Improvement, and External Tool Usage.
• Figure 4: Two frameworks towards LLM-based educational framework. (a) depicts the unified approach, where a single LLM addresses all aspects of educational-related queries, utilizing its internal capabilities such as mathematics, writing, knowledge-based question answering, reasoning, and programming. (b) illustrates the Mixture of Experts (MoE) approach, where an LLM controller is tasked with task distribution, delegating specific questions to specialized expert models that are proficient in individual areas.