> For the complete documentation index, see [llms.txt](https://examind.gitbook.io/v1/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://examind.gitbook.io/v1/feedback-machines/bulk-grading-assist.md).

# Bulk Grading Assist

Grade a whole class at once — upload a batch of submissions, let the machine evaluate them, and export the results.

Bulk Grading Assist lets you grade a whole class's work at once. You upload a batch of student submissions, the Feedback Machine evaluates each one against your criteria, and you review and export the scores and feedback.

{% hint style="info" %}
**Bulk Import** and **Bulk Export** are available to anyone who can manage the machine — its owner, and instructors in its class. They're no longer enabled per institution or class. If you don't see them on a machine, check that you're an instructor in the class it's shared with, or [contact us](mailto:support@examind.io).
{% endhint %}

## Grade a batch

{% stepper %}
{% step %}

### Start a bulk import

From your Feedback Machine's menu, select **Bulk Import**. Upload a **`.zip` containing your students' `.docx` and/or `.pdf` files** — each file becomes its own submission (only `.docx` and `.pdf` files are processed). A single import accepts up to **1 GiB** of files (with individual files up to 30 MB). Then select **Start Bulk Import**, and keep the browser tab open until all submissions are created.

{% hint style="info" %}
In Canvas, an assignment's **Download Submissions** option gives you a zip of student files you can upload directly. Feedback Machines recognizes the Canvas filename format and groups multiple files from the same student.
{% endhint %}
{% endstep %}

{% step %}

### Let it evaluate

Feedback Machines creates a submission for each file and evaluates it against your criteria. Evaluations run in the background — **up to 200 at a time** — so large batches process steadily without you waiting on each one; when the system isn't busy, a full class of around 1,000 submissions can finish in as little as an hour. You'll see live progress for each submission; any that fail are retried automatically, and you can retry remaining errors yourself.
{% endstep %}

{% step %}

### Review results

The import lists every submission with its score and status. Open any one to read its full feedback and criteria breakdown — including anything hidden from students. To read the batch by rubric part rather than one submission at a time, and to approve the grades, use the machine's [Results page](/v1/feedback-machines/review-adjust-and-approve.md).
{% endstep %}

{% step %}

### Export

Select **Export Results** to download:

* **Canvas gradebook** — scores as a CSV you can import into Canvas.
* **Feedback files** — a zip of per-student feedback.
* **Submissions** — a zip of the original student files.

You can export all students, or only each student's highest-scoring submission. The detailed CSV export includes separate **First Name** and **Last Name** columns alongside each student's email.
{% endstep %}
{% endstepper %}

## Refine and re-evaluate

Bulk grading is iterative. As you review the results, you'll often spot evaluations you'd have graded differently — that's expected, and bringing the machine into agreement with your judgment is the core of the workflow. A few rounds is normal.

The place to do it is the machine's [Results page](/v1/feedback-machines/review-adjust-and-approve.md). There you can read the whole batch by rubric part, pull the submissions you disagree with straight into the **Adjust** chat, preview the corrected evaluations, and apply the change across every submission at once — then approve. See [Review, Adjust & Approve Results](/v1/feedback-machines/review-adjust-and-approve.md).

{% hint style="info" %}
You can also select **Re-evaluate All** from the import to re-run every submission against the current machine — useful after editing the machine directly in the [Modify panel](/v1/feedback-machines/build-a-feedback-machine/modifying-a-feedback-machine.md). Your original import and its results are preserved either way, so you can compare before and after.
{% endhint %}