OBE-Based-Exam-Paper-Evaluation

Multimodal Examination Answer Dataset with Expert-Designed OBE Rubrics for Criterion-Level Assessment

Cite this work if you use this dataset

@misc{sm2026multimodalexaminationanswerdata,
      title={Multimodal examination answer data with expert-designed Outcome-Based Education rubrics for criterion-level assessment}, 
      author={Jahangir Alam SM and Md Khalid Syfullah and Saad Ahmed and Munira Akter Mou and A K Z Rasel Rahman and A. K. M. Masudur Rahman and Mohammed Sowket Ali},
      year={2026},
      eprint={2608.22346},
      archivePrefix={arXiv},
      primaryClass={cs.CV},
      url={https://arxiv.org/abs/2608.22346}, 
}

A multimodal collection of 485 scanned examination answers from 415 consenting students across four institutions, contributed by eight faculty members and covering nine subjects and 12 question templates (47 rubric criteria).

Each answer pairs a scanned PDF with a JSON record containing the subject, question, model answer, Outcome-Based Education (OBE) rubric criteria, performance-level descriptions, criterion-level marks, and total mark. Scans retain authentic content — handwriting, printed text, equations, tables, code, figures, and diagrams — captured via CamScanner, Adobe Scan, and conventional scanners, with natural variation in illumination, orientation, resolution, and revision marks.

Uses: rubric-aware automated evaluation, multimodal document understanding, OCR, feedback generation, score prediction, and privacy-aware OBE assessment.

Access & ethics: Restricted to research use, available upon reasonable request. Student consent was obtained; the dataset was kept separate from official grading. Some scans may retain residual identifiers — recipients must avoid re-identification and redistribution, retain human oversight, and not treat model outputs as official grades.

Funding: Bangladesh Accreditation Council (BAC), project AI-Powered Automated Exam Evaluation and OBE-Based University Assessment System.