TA Grading Rubric
Defines a 50-point grading rubric with four categories and a core principle of leniency for demonstrated understanding.
What this file does
Defines a 50-point grading rubric with four categories and a core principle of leniency for demonstrated understanding.
When to use it
- Standardising grading across multiple TAs for programming assignments
- Creating a rubric that prioritises conceptual mastery over minor errors
- Providing consistent feedback with point deductions and constructive notes
- Scaling grading from small to large courses with multiple sections
TA Grading Rubric
Total: 50 Points
Core Principle: Focus on whether the student has demonstrated mastery of the underlying concepts. If a minor detail is missed but competence is clearly shown elsewhere, apply minimal or no deduction. Deduct heavily only when the student clearly has not grasped the core concept.
Score Target: Most students should score 45-50. The floor is 40+. Be generous — prefer giving the benefit of the doubt. Only deduct when there's a clear, meaningful gap in understanding. Small issues that don't reflect a lack of skill should cost at most 1 point, or nothing at all.
I. Workflow
- Code Review — Inspect logic, structure, code quality, and technical requirements defined in the assignment description.
- Output Review — Verify output against expected results, assess correctness and usability.
II. Rubric Categories
1. Follows Directions (5 pts)
Compliance with administrative, structural, and submission requirements.
| Issue | Deduction | Notes |
|---|---|---|
| Naming violation | -1 to -2 | Files/folders don't follow the specified naming convention. If the intent is clear and everything works, lean toward -1. |
| Documentation issues | -2 to -3 | Missing or incomplete README that prevents smooth setup/execution. |
| Submission error | -5 | Wrong platform, invalid repo link, or inaccessible work. |
2. Quality (15 pts)
Code robustness, output accuracy, and professional engineering standards.
| Issue | Deduction | Notes |
|---|---|---|
| Critical failure | -10 to -15 | Program crashes on launch or core functionality consistently fails. |
| Output mismatch | -3 to -8 | Output deviates significantly from the expected results or specification. Minor differences are acceptable; missing entire required components is a functional gap. |
| Code fragility | -2 to -5 | Lack of basic error handling, poor structure, or fragile logic. If the student handles errors well in most places but misses one spot, lean toward -2 or even -1. |
3. Required Function (25 pts)
Completion and correctness of the assignment's core engineering objectives.
| Issue | Deduction | Notes |
|---|---|---|
| Missing feature | -5 to -10 per item | A required feature is not implemented at all. |
| Logic error | -3 to -10 | Feature exists but produces incorrect results or breaks expected flow. If the approach is sound but has a small bug, lean toward -3. |
| Hardcoding | -5 | Static values where dynamic computation is required. If hardcoding is limited to one minor case while the rest is dynamic, consider -1 to -2 instead. |
4. Timeliness (5 pts)
| Condition | Deduction |
|---|---|
| On time | 0 |
| Grace period | -1 to -2 |
| Late | -1 per day, up to -5 |
III. Grading Guidelines
- Demonstrated mastery > surface perfection. A student who clearly understands the core concepts but has a minor naming inconsistency should not be penalized the same as one who doesn't understand the concepts at all.
- Context matters. Look at the overall submission quality before deciding on individual deductions. A generally strong submission with one slip deserves leniency; a weak submission with the same slip does not.
- Be consistent with prior assignments. Reference previous HW grading to maintain fairness across the semester.
- Feedback should be constructive. Always explain why points were deducted and suggest how to improve.
IV. Output Format
For each student, provide:
- A list of concrete issues found
- Point deductions aligned with this rubric
- Clear, constructive feedback
What's inside
4 rubric categories, 12 issue types with deduction ranges, grading guidelines, and output format instructions.
Change this for your project
- Replace
TA Grading Rubricwith your course or assignment name - Adjust point totals like
Total: 50 Pointsto match your assignment weight - Modify
Score Target: Most students should score 45-50to your desired grade distribution - Update
Reference previous HW gradingto reference your own prior assignments
Where it goes
Keep it in your repository where the agent or team that needs it will read it.
Worth borrowing
- Core principle statement that sets grading philosophy upfront
- Issue table with deduction ranges and contextual notes for each row
- Grading guidelines that reinforce consistency and constructive feedback
Related Documents
AICA PRD Quality Rubric
Defines a 10-dimension scoring rubric (0-5 each) to evaluate and create PRDs, with a target score of 36+ for agent-driven implementation.
Judging Rubric
Defines a 100-point scoring rubric for an AI-for-social-good hackathon, covering five criteria with detailed score bands and tie-breaking rules.
Liscribe v2 — Rubric of Success
Defines the complete UI specification, success criteria, and non-negotiable principles for a macOS transcription app with three independent workflows.
QualRubric
Defines a qualitative rubric for PhD qualifying exam presentations, focusing on context, organization, and knowledge demonstration rather than numerical scoring.