Students reported first-pass false-positive deductions at a mean of 2.94 per term, normalising to about 0.6% of rubric items¶

Claim · All claims
Evidence · 1 study · design r2 · q2
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person

Subclaims¶

q2 i? Students reported a mean of 2.94 (SD = 2.99, median = 2) first-pass false-positive deductions over the term, a per-rubric-item rate of approximately 0.6%, versus ≈0.22% for missed deductions. → Jesse A. Rodríguez 2026

Evidence¶

Jesse A. Rodríguez 2026¶

Jesse A. Rodríguez. (2026). LaTA: A Drop-In, FERPA-Compliant Local-LLM Autograder for Upper-Division STEM Coursework. arXiv preprint. https://arxiv.org/abs/2605.05410 doi:10.48550/arxiv.2605.05410

q2 · i? · design · r2

Block 2 of the anonymous post-term survey asked students to count grading errors observed over the term; counts were capped and verified against free-text items. The quote reports the printed first-pass false-positive statistics.

"Across the entire term, students reported a mean of 2.94 (SD = 2.99, median = 2) first-pass false-positive deductions (instances where LaTA took points off incorrectly) with a range of[0,20] ."

Discussion¶

Related Claims¶

• Instructor-confirmed grading errors occurred at roughly 0.02–0.04% of rubric line items across the term — related
• An on-premises LLM autograder graded a 200-student course at 1–3 minutes wall-clock time per submission across a full term — related
• Student perceptions of the autograder were mixed, with the LaTeX workflow rated more positively than the grader itself — related
• GPT models showed lower accuracy with more rubric-related errors (28% vs. 13% for Gemini 3 Flash) and near parity between false positives and negatives — related
• The TA ground truth was produced under a one-rater-per-item design with co-grading, an exceptions log, instructor audit, and student review, with an estimated grading-error rate of about 0.1% in the student-inspected subsample — related