LaTA: A Drop-in, FERPA-Compliant Local-LLM Autograder for Upper-Division STEM Coursework RodrΓguez (2026) β Oregon State University. Submitted to Computers & Education. π Full text (arXiv)
Summary
LaTA (LaTeX Teaching Assistant) is a privacy-preserving, drop-in autograder that addresses the core tension in educational LLM deployment: most automated-grading systems send student work to third-party APIs, violating FERPA and exposing institutions to data risk.
System architecture:
- Four-stage pipeline: Ingest β Segment β Grade β Report
- Grading engine: gpt-oss:120b running entirely on local commodity hardware (single Mac Studio)
- Rubric format: YAML with binary per-item scoring, comparing student work against instructor-authored reference solutions
- Cost: $0 marginal cost per assignment, 1-3 minutes per submission
Real-world deployment (Winter 2026):
- Course: ME 373 at Oregon State University (~200 students)
- Grading error rate: 0.02-0.04% per rubric line item β near-perfect accuracy
- Operational benefit: Enabled regrading of corrected assignments and expanded TA office hours
Learning outcomes (vs. traditional cohort):
- Midterm exam: +11% performance gain
- Final exam: +8% performance gain
- Confidence gains: Ξ β₯ +1.49 Likert points on every learning objective (N=159, p < 10β»Β²β·)
These results provide strong evidence for the feedback-loop hypothesis: faster, more consistent feedback drives both learning and confidence. The deployment demonstrates that generative-ai grading can be both FERPA-compliant and pedagogically effective, addressing concerns raised in assessment-validity and formative-assessment discussions.
LaTA's success connects to the broader stem-education and higher-ed landscape, showing that llm-based grading can move beyond short-answer-scoring-quality-degradation concerns when properly designed with instructor-authored rubrics and reference solutions. The open-source, zero-marginal-cost model aligns with principled-ai-education principles.
Related Pages
- llm-handwritten-math-grading β Complements LaTA's typed-response grading with vision-based handwritten math assessment
- automated-grading β Automated grading systems
- automatic-short-answer-grading β Short answer grading with LLMs
- formative-assessment β Formative assessment practices
- feedback-loop β Feedback mechanisms in learning
- assessment-validity β Assessment validity in AI education
- stem-education β STEM education and AI
- principled-ai-education β Principled AI in education
- short-answer-scoring-quality-degradation β Quality issues in automated scoring