Vision LLMs achieve high rubric-level accuracy on handwritten math but most errors stem from transcription failures rather than rubric misapplication.
In: Proceedings of the 57th ACM Technical Symposium on Computer Science Education V.1
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CY 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Automated Grading of Handwritten Mathematics Using Vision-Capable LLMs
Vision LLMs achieve high rubric-level accuracy on handwritten math but most errors stem from transcription failures rather than rubric misapplication.