Cropping scanned exam pages to a question's bounding box improved handwritten-answer grading accuracy and cut token/compute cost across eight vision-language models.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Bounding Boxes to Improve Small Language Model Performance on Vision-Based Grading Tasks
Cropping scanned exam pages to a question's bounding box improved handwritten-answer grading accuracy and cut token/compute cost across eight vision-language models.