Step-by-step workflow
How to use LLMGrader
Follow one accounting example from provider setup and quiz creation through instructor-reviewed grading, student feedback files, analysis of learning, and progress over time.
Short, task-focused demonstrations
Video guide chapters
The videos will use the same accounting example throughout. Quiz-creation chapters will use the demo questions, solutions, and rubrics as examples; importing and grading chapters will use the demo answers for the accounting quiz.
- 1Video chapterSet up an AI provider
- 2Video chapterCreate a new quiz
- 3Video chapterUse the Chat Panel while building a quiz
- 4Video chapterImport student answers
- 5Video chapterGrade the accounting demo answers
- 6Video chapterGuide another grading pass
- 7Video chapterCreate password-protected PDF feedback files
- 8Video chapterUse Grade Summary and Common Theme Analysis
- 9Video chapterExplore the accounting Course Progress demo
- 10Video chapterSave work and export results
Choose and test an AI provider
Open Settings and select the model source appropriate for the course. Desktop users can run a local Ollama model, use Ollama Cloud, or connect to OpenRouter, OpenAI, Anthropic, or a compatible custom endpoint. Browser users can use a supported cloud provider or an Ollama-compatible institutional endpoint.
Enter any required API key, select Refresh Models / Check Provider, choose a model, and save the LLM choice in preferences.
Create the accounting quiz
Choose Create a new quiz. Enter the course and quiz information, the number of questions, and the points available for each question.
The guide will use questions from the built-in accounting demo as concrete examples. Desktop users also select or create the course folder where saved assessments will be organized. LLMGrader checks that the folder does not contain assessments from a different course or section.
Add questions, solutions, and rubrics
For each question, enter the exact wording students received, the solution or expected answer, the available points, and any question-specific rubric.
The optional Chat Panel can help draft or refine a question, verify a solution, create a rubric, and generate examples of stronger and weaker answers. The instructor reviews anything produced by the model. The video example will use the question, solution, and rubric from the accounting demo.
Import the accounting demo answers
Import answer data after the quiz structure is complete. LLMGrader provides a guided two-step D2L workflow as well as its own CSV import and export format.
The guide will use the built-in demo answers for the accounting quiz, keeping the same fictional students and responses throughout the remaining chapters. Review the imported student and question counts before grading. One grading operation is limited to 1,000 answers to avoid unintentionally placing excessive demand on a local computer or provider account.
Grade and review results
Grade one answer, one question, or the remaining quiz. LLMGrader sends the question, solution, rubric, relevant prompts, and anonymized student response to the selected model.
Using the accounting demo answers makes it possible to show navigation between students and questions, the progress messages shown during grading, and the instructor review of each suggested grade and comment.
Use Grading Guidance
When a suggested grade or comment needs reconsideration, use Grading Guidance to explain what the model should examine in another pass. The guidance supplements the existing question, solution, and rubric rather than replacing the instructor’s grading standard.
Review the revised grade and feedback, edit either field directly if needed, and save only the instructor-approved result.
Create student feedback files
LLMGrader can create a separate PDF feedback document for each student. Each PDF can include the question, solution, student answer, grade, and instructor-approved feedback.
The individual PDFs can be password protected. LLMGrader displays the generated student passwords, creates a ZIP containing the feedback files, and can include the passwords in the supported D2L workflow.
Review the class and identify common themes
Open Grade Summary to compare total and question-level performance. Select individual students or show the whole class when a broader comparison is useful.
After grading is complete, run Common Theme Analysis. It identifies recurring strengths, misconceptions, incomplete explanations, and possible areas for instructional follow-up across the class.
Examine progress over time
In the desktop version, Course Progress reads compatible saved quizzes from a selected course folder. It shows student and class grade trends and can compare performance across assessment dates.
The Course Progress video will use the five-assessment accounting demo. Conceptual analysis uses the Common Theme results from the saved quizzes, and LLMGrader can create missing analyses after confirming that the underlying answers have been graded.
Save and export
Save the quiz record before closing. Depending on the workflow, LLMGrader can produce its own CSV files and D2L-compatible grade files in addition to the individual student feedback documents.
Application data remains under the instructor’s control. Browser files are explicitly downloaded or selected by the user; desktop course records are stored in the chosen course folder.
If something does not proceed
No models appear: verify the endpoint or provider, API key if required, and network access, then refresh the model list.
Grading cannot start: check that the question, solution, point value, rubric, and student answer are present.
Course Progress does not match quizzes: use one folder containing assessments for the same course, term, section, and student group.
A browser feature cannot access a file or clipboard: follow the browser permission prompt or use the application’s file-selection and Paste controls.