TimelyGrader October Update

Section and TA support, submission visibility, regenerate feedback, grading explanations, and GPT-5 update.

Some long awaited features are coming this Thursday, October 16th, at 4PM PST to our self serve and API users. Make your feedback sound more like you or Deadpool?

Alright, buckle up buttercup: here comes the list of updates you didn’t ask for but totally need.

1) Section and TA support (API)

For large enrolment courses, complexity increases the more teaching assistants and sections there are. Not to mention controlling visibility so that instructors and TAs are seeing only student data in their specific sections.

With this new update, we have created a new role on TimelyGrader, teaching assistants (TAs). They are automatically assigned when they open TimelyGrader as we can retrieve their roles from the LMS. We are also retrieving section data for each submission. More on this later:

TAs have section data attached

2) Limiting visibility based on sections (API)

TAs now see only the sections they’re assigned, along with the relevant students, submissions, and analytics. This prevents any cross section privacy and visibility issues that may arise from large enrolment courses.

As a result, if sections are assigned to TAs on the LMS, their visibility will also be limited on TimelyGraer so they will only grade the student submissions in their respective sections.

TAs can only see student submissions within their assigned sections

3) Adding your own grading explanations (Self-serve & API)

Sometimes we may suggest a grade and feedback that you disagree with (surprise surprise, hallucinations can happen). This often meant that the instructor needed to manually adjust the feedback if they did not agree with the grade. We understand that this can be a time consuming process.

To remedy this, instructors can now add their grading explanation so they can leave a sentence to explain why this criterion deserves a higher or lower score. Even if we suggested the correct rating, there might be a different reason why the student deserves this rating. More on how we make use of this data later:

4) Regenerating feedback based on instructor grading explanations (Self-serve & API)

Naturally, we can make use of the instructor grading explanation and regenerate the feedback for the entire submission so that the feedback is more relevant to the actual grade. This eliminates the need to manually write feedback from scratch.

Essentially, TimelyGrader will take a first pass at the feedback. If you think the feedback can be better, you can now regenerate the feedback for each submission so that it is more relevant. You can refine your feedback so it also sounds more feedback you would give.

5) Updated “View Submissions” page (Self-serve & API)

The submissions table now does more with a nuanced layout, improved filters, multi-select, and pronounced buttons (aka we took the buttons out of the ellipsis). You can still Push Grades/Feedback to your LMS and reset your grading.

Note that for TAs and our API users, they do not have the option to reset and push grades. That would be the job of the instructor.

6) GPT-5: Everyone gets a faster engine!

Over the past month, we have received numerous reports of generation times taking a significantly longer time than expected. This was the result of a queuing system we placed due to the significant uptick in usage. We went from 4 institutions to 25 institutions and counting!

With the GPT-5 update and because of other improvements from our own R&D, generation times are noticeably quicker (up to 70%). Feedback is also more context-aware across the entire rubric, instructor explanations, and it sounds much more human-like.

You want to grade quizzes, exams, problem sets, and more? Coming soon to TimelyGrader. ;)