Skip to content

LLM-embedding autograding pipeline for freeform mathematical proofs

Research Method · All research methods
Evidence · no claims cited

Description

The paper describes an automated grading process in which a student-written proof is first run through a pretrained large language model to obtain embeddings, and a custom classification model is then applied to those embeddings to calculate accuracy on various rubric points. Embedding selection and grading-model fitting are decoupled so the expensive semantic feature extraction is reused and a simple classifier (e.g., linear regression) suffices for the downstream grading task.

Accounts

  • LLM-embedding autograding pipeline for freeform mathematical proofs: The paper describes an automated grading process in which a student-written proof is first run through a pretrained large language model to obtain embeddings, and a custom classification model is then applied to those embeddings to calculate accuracy on various rubric points. Embedding selection and grading-model fitting are decoupled so the expensive semantic feature extraction is reused and a simple classifier (e.g., linear regression) suffices for the downstream grading task. (Zhao et al. (2025))

Claims

-

Key Sources

  • Zhao, C., Silva, M., & Poulsen, S. (2025). Autograding Mathematical Induction Proofs with Natural Language Processing. International Journal of Artificial Intelligence in Education, 35, 3202–3232. https://doi.org/10.1007/s40593-025-00498-2
Page metadata
Field Value
type research-method
id llm-embedding-autograding-pipeline-for-freeform-mathematical-proofs
title LLM-embedding autograding pipeline for freeform mathematical proofs
description The paper describes an automated grading process in which a student-written proof is first run through a pretrained large language model to obtain embeddings, and a custom classification model is then applied to those embeddings to calculate accuracy on various rubric points.
status review
generated by process:settle-candidates on 2026-10-27
trust unverified — no one has confirmed this page's content

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