Ednius uses a multi-agent AI system to keep grading accurate and free from AI hallucinations—when an AI makes up details. Here's how it works:
Generative Agent: It starts by reviewing student work and giving an initial grade with feedback. Sometimes it may add details that aren’t in the data.
Verification Agent: It double-checks the generative agent’s output for any made-up or incorrect information.
Orchestrator Agent: This agent manages the process. If any issues are flagged, it triggers another review to correct errors.
Rule-Based Agents: They handle simple tasks like checking typos and verifying numbers, adding another layer of accuracy.
This system minimizes errors, ensures clear grading criteria, and gives both educators and students reliable results.