Skip to main content
Assignment Evaluation marks everyday class work. Your platform uploads the worksheet or question set, the AI reads out the questions, your platform supplies either model answers or a rubric, and then each student’s work is scored question by question. It is the lighter of the two evaluation products. There is no approval step and no publishing stage, so scores are readable as soon as the AI finishes. Nothing in the API checks who is calling, so your platform decides which of its users may set up an assignment and who may read a score.

​
What can an assignment be marked against?

One of two things, and you pick when you set the assignment up.

Model answers

Upload the answer key as a PDF or PNG photographs. The AI reads the expected answer for each question, with the marks for each step of it, so a student who gets the method right but the answer wrong can still earn part of the marks.

A rubric

Attach a rubric instead of an answer key. A rubric is a list of criteria, each with a score range and a description of what earns each band, for example “4 to 5: explains the idea in their own words with an example”.
A rubric suits open work where there is no single right answer, such as a spoken explanation, an essay, or a presentation. Model answers suit work with definite answers.

​
Which subjects can an assignment cover?

There is no subject setting, and this is the product’s biggest limit. Every assignment goes to the marking AI with a fixed subject and curriculum, English on the NCERT syllabus, whatever the work actually is. An assignment record has no subject field at all, so sending one changes nothing. The grade is the one class detail that is passed through. If you leave it blank, the work is marked as grade 10.
Do not read an assignment score as a subject-aware mark. On a maths or science assignment with model answers, the AI is still told it is marking English, so it judges the writing as much as the subject content. Check a sample before you put assignment marking in front of teachers for a subject other than English.

​
What can students submit?

You cannot mix a PDF with images on the same answer sheet, and you cannot send files and an audio link in the same request.
Send PDF or PNG, not JPEG. A JPEG upload is accepted, but the assignment side does not record it as an image, so the checks that stop you mixing a PDF with photographs stop working and later uploads on the same item can behave strangely. Convert photographs to PNG, or put them in one PDF. Word files are not supported here at all.
An audio answer only works if the assignment’s model answer sheet has a rubric attached. Without one the API refuses it, because there is nothing to score the recording against.

​
What comes back for each question?

The whole assignment is returned in one response, grouped by question, so you get the student’s answer next to the question and the expected answer.

​
How does a rubric work?

You build a rubric once and reuse it across assignments.
The band wording is not just documentation. The AI reads it to work out what the criterion is measuring, so rewording a band can quietly change how work is scored. Review a rubric before reusing it on live work, and keep changes deliberate.
Criteria can also be created on their own and then linked to several rubrics, which saves rewriting a criterion such as “spoken fluency” for every subject.

​
What are the limits?

Assignment Evaluation does not accept video, and it does not identify who is speaking in a recording.

​
What does it not do that Exam Evaluation does?

Worth knowing before you choose:

​
Frequently asked questions

You attach one or the other. If you attach a rubric, the rubric is what the AI scores against.
Not through the API. Assignment Evaluation has no endpoint for editing a score and no approval flag. If a teacher has to sign marks off, use Exam Evaluation instead.
No. Audio only, as an MP3 file at a public HTTPS address. Convert the recording to MP3 and host it somewhere reachable before you send the link.
Once an assignment has been processed, sending it for processing again returns an error saying it is already extracted. Create a new assignment if the question set changes.
Both the assignment and its model answer sheet have to be fully processed first. Until then, creating an answer sheet returns an error saying the assignment or the model answer sheet is not extracted yet.

​
Next steps

How it works

The full flow from question set to scores.

Use cases

How schools and platforms use assignment evaluation.

Developer guide

Evaluate an assignment with the API, step by step.

Create a rubric

Build a rubric and its criteria with the API.