Mahal Systems EduBench
Fits education assistants that serve both students and teachers. Borrow its nine-scenario taxonomy and the 12-indicator scoring sheet (scenario fit, factual accuracy, pedagogy), with human raters used to check the model grader.
What it asks the agent to do
- answer a student's question
- correct a student's error
- give personalized learning support or emotional support
- generate quiz questions or teaching material
- grade a student's answer
What to borrow from it
taxonomygrading design
Caveats
All prompts are LLM-synthesized, not real classroom data; scoring is LLM-judge on a 1 to 10 scale with modest human agreement (Kendall's W around 0.6); single-turn only.
Details
| Workloads | education, knowledge qa, documents |
|---|---|
| Interaction | single turn |
| Environment | static |
| End state checked | No |
| Tools | none |
| Grading | llm judge, human |
| Size | 4,000 |
| Data | synthetic |
| Personal data | none |
| License | MIT |
| Harness | own repo |
| Status | active |