AI Grading for CBSE/ICSE Worksheets: How Report Cards Work
A worksheet is only useful if you know what you got wrong and why. Here is how Mingi turns every attempt into a report card parents can actually use.
A worksheet is only useful if you know what you got wrong and why.
Mingi turns every attempt into a report card at /report/[id] and updates topic mastery from real scores.
An unmarked worksheet is theatre. You spent forty minutes, felt busy, and still do not know whether Electricity or careless arithmetic caused the drop. The product that matters is the report card: per-question marks, why it was wrong, and which topic to retry.
On Mingi, every submitted worksheet attempt becomes a report at /report/[id]. Grading uses the grade-exam path for MCQs and written sections. Grading does not spend your generation quota — generations create papers; grading scores them. Scores update learnability metrics so topics below 60% mastery surface on student and PIN-gated parent views.
Who this is for
- CBSE and ICSE students who finish papers but never review wrongs systematically
- Parents who want formative evidence without re-teaching every chapter
- Families comparing chat homework helpers (no marks) with graded practice loops
Formative report card: A Mingi report card is practice feedback: marks per item, rubric-style comments, and topic tags for the next drill. It is not a CBSE/ICSE board marksheet and not a school report card replacement. Use it to decide what to retry in 48 hours.
/report
Report card path
0 gen
Grading quota cost
60%
Weak mastery flag
48h
Retry gap
Solve → grade → mastery loop
Generate or open a worksheet, solve interactively, submit, then open the report URL. Read per-question feedback before you peek at full model answers — that is where you learn what the examiner-style comment is pointing at. Then write one sentence: the single idea that still confuses you.
If you only glance at the total percentage and close the tab, you have used AI as a score machine, not a teacher. The mastery bars only move when graded items land with topic tags — which means finish, submit, and spend five minutes on the lowest two questions before you open a game or the next PDF.
From attempt to next drill
Submit the paper
Unfinished attempts do not update mastery — finish or intentionally pause with a plan.
Open /report/[id]
One URL per graded attempt; bookmark or share with parent for Sunday review.
Read item feedback first
Then model answers. Rewrite the two worst items with full working.
Check mastery flags
Topics under ~60% become the next 20-mark generation target.
Retry in 48 hours
Fresh paper on the same topic — same-evening identical sets overestimate learning.
| Use the report for | Do not use it for |
|---|---|
| Naming weak sub-topics after practice | Official board or school final marks |
| Parent Sunday one-misconception review | Replacing teacher feedback on long essays |
| Choosing tomorrow’s 20-mark chapter | Comparing children publicly by percentage |
| Seeing whether retries improved mastery | Skipping NCERT / specimen papers entirely |
Parents: share the report card weekly, not every night. One misconception from the feedback is enough. “What did question four say you missed?” beats “Why is your total only 62?” Custom Test and standard worksheets both land on the same grading and mastery pipeline when submitted.
Quota, long answers, and honest limits
Creating a new paper spends an AI generation (trial 15, Starter 60/month). Submitting for grading does not spend a generation. Skipping the report to “save” quota is the opposite of how the product creates value — you already spent the generation on an unreviewed paper.
Generation vs grading quota: Always grade finished attempts. Mastery flags and parent dashboard bars only update from real scores, not from half-open worksheets left in activity history.
Borderline long answers
AI awards formative marks using rubric-style feedback. Borderline written answers may still deserve a teacher’s eye. Treat AI as a fast first pass that tells you where to ask the human — not as a court of final appeal for boards.
⚡ Parent move: Say: "Open last night’s report link and show me one question’s feedback. We retry that topic Wednesday — I am not re-teaching the whole chapter."
For the student reading this: Before you generate another paper, open the last report and rewrite the two lowest-scoring items. That is where marks return.
- Report cards at /report/[id] per graded attempt.
- grade-exam covers MCQ + written; grading free of generation quota.
- learnability_metrics from real scores; <60% topics flagged.
- Starter ₹1,299/mo · 60 gens; trial 30 days · 15 gens.
Know it sooner
Know it sooner which sub-topic leaked marks before the next class.
Solve a 20-mark paper and open your report cardHonest limits
- AI grading is formative practice feedback, not official board marksheet replacement.
- Borderline long-answer marks may need teacher review; use AI feedback as a first pass.
- Report cards only help if you finish and submit — abandoned papers create no mastery signal.
See this week without asking.
Know it sooner which sub-topic leaked marks before the next class.
Open the parent dashboardThe Takeaway
- Read per-question feedback before model answers; that is where marks are won back.
- Share the report card weekly with parents; one misconception per review session.
- Retry weak topics after 48 hours once mastery data updates.
- Use AI grading for practice volume; keep tuition for new concepts and deep doubts.