Evaluation of a User Interface Extension Integrating an LLM-Based AI Assistant into an Interactive Visual Equation Editor for Solving High School Mathematics Problems
Year: 2026
Type: article
Abstract: This paper presents an evaluation of AI-assisted human–computer interaction for mathematical problem solving within a multimodal mathematical editing environment. The proposed architecture integrates a visual equation editor, voice-based interaction, REST communication services, and externally hosted generative AI large language models (LLMs) to support the creation, interpretation, and solution of mathematical expressions. The study was conducted using the Equation Wizard environment, which provides multimodal mathematical content editing based on both proprietary and standard formula representations, including MathML and LATEX. A conversational AI interaction layer enables users to communicate with selected LLMs using natural language voice commands. The main objective of the study was to determine whether contemporary LLM-based services can reliably support mathematical problem solving within an AI-enhanced equation editing environment. To address this objective, seven contemporary GenAI LLMs were evaluated using a benchmark consisting of representative high school mathematics problems covering algebra, limits, trigonometry, logarithms, inequalities, and parameterized expressions. The evaluation focused on mathematical correctness, output interpretability and visualization quality, response latency, and compliance with mathematical encoding formats within the complete interaction workflow. The study also compares representative model families with respect to correctness, syntactic compliance, and responsiveness within the complete interaction workflow. Unlike conventional LLM benchmarks that assess models in isolation, this work evaluates end-to-end AI-assisted mathematical interaction involving the editor, communication infrastructure, and language models. The study demonstrates a practical approach to assessing the usefulness of multimodal AI-enhanced mathematical problem solving in realistic usage scenarios. The results show that, within the evaluated interaction workflow, current LLMs achieve high levels of mathematical problem-solving performance, although substantial differences were observed in response latency and output-format compliance. An important finding is the discrepancy between the models’ strong semantic understanding of custom-encoded mathematical input and their weaker ability to generate syntactically compliant encoded outputs, highlighting a key challenge in the integration of LLMs with structured mathematical software environments.
Source: Electronics
Authors: Agnieszka Bier, Zdzisław Sroczyński
Institution: Silesian University of Technology
Language: English
Subfield: Computational Theory and Mathematics
Field: Computer Science
Domain: Physical Sciences
SDG: Quality education
Open Access status: gold
Estimated APC paid: $2,896
OpenAlex ID: W7168692790
FWCI:
Cites:
Cited by:
Related to: