Computer Science Assessment Stack

Enter to continue.

No personal information is needed or recorded.

Computer Science Assessment Stack

Ingest · Generate · Exam · Deploy · Grade — grounded in your material, calibrated to how you grade.

checking models…

Upload material → it's chunked, embedded, and saved as a reusable RAG index. Everything downstream grounds on this.

Step 1 — build a knowledge base from your material.
Advanced RAG settings (improve retrieval quality)
Smaller chunks = more precise retrieval; overlap keeps ideas from being split. Embedding model is set in config.py.

Every RAG is saved with a name starting with rag_ (so it's easy to spot in lists).

none yet

Pick material, generate grounded questions, then name & save the set.

none yet

Build an exam from your saved question sets (made in Q-Modeler). Name it, tick the questions you want (none selected by default), and create — it becomes available in Deploy Exam.

📚 Question Bank

Your saved question sets. Expand a set to tick the questions to include; delete a set with 🗑.

Deploy an exam to make it available to students. Share the student link; students pick a deployed exam, enter their name, and take it.

Student link: /student  

Every exam students submitted. Grade one with a saved Avatar (built in Clone Lab) — if you pick none, the default grader is used. After grading, a report link appears.

Your Grader DNA. The building blocks of how you grade — Behaviors, your Grader Personality, and Rubrics. Define them here, then in Time to Clone You! we find the mix that grades most like you. A person has one personality, so pick just one when you calibrate.

Behaviors

How the grader should behave — tone, strictness, what to emphasize. Take the quick quiz to draft one.

Grader personality

Learned from a few answers you already graded, or built from a quick quiz.

✨ More advanced training (learning from your own graded examples) is coming soon.

Rubrics

Criteria for scoring.

Default (used if none chosen):

🧬 Time to Clone You!

Find the AI "grader" that scores most like you — a mix of your Behavior × Personality × Rubric.

You ≈ argmin (B × P × R)    Σi   |   gradeAI( B, P, R · answeri )  −  gradeYou( answeri )   |

In plain words: you pick a combination of Behaviors × Personalities × Rubrics, then show your grading style by scoring a question's answers yourself. On Clone Me!, the tool grades those same answers with each combination through the LLM and compares to your scores — the combination that disagrees with you least is your closest clone. More graded answers → a sharper clone.

① Pick the ingredients to test

Each combination of Behavior × one Personality × Rubric is one candidate "you." Tick any behaviors and rubrics; pick a single personality (a person has one). Max 12 combinations.

Behaviors (pick any)
Personality (pick one)
Rubrics (any; none = default 0–5)
② Let's learn you!

Show the AI how you grade: get a question, then score a spread of answers yourself.

🎲 Option A — Auto-generate a question from a RAG
Generates a grounded question + a spread of student answers right here for you to grade.
✍️ Option B — Bring your own question & grade the answers
🧬 Clone Me! compares each combo's grades to yours and ranks them. 🏅 Optimal Avatar does the same, then saves the closest‑matching combo as an avatar. 💾 saves the one combo you've ticked.