Release Notes

Improved: AI Review Now References Your Scoring Guide

When a rubric includes a scoring guide, the AI reviewer will now explicitly use that guidance when assigning ratings and will reference the relevant portions of the scoring guide in its explanation. Previously, the scoring guide content was available to the AI but it was not instructed to cite it, which meant rating rationales could lack the connection to the rubric author's intended evaluation criteria. This improvement applies to single-criterion reviews, full multi-criterion reviews, and human review AI analysis.

Improved: Clear Error Messages When a Model Is Unavailable

If the selected AI model has been deprecated or removed by its provider, the system now displays a clear message explaining what happened and advising you to select a different model. Previously, this situation could result in an empty review with no explanation. The system also now stops immediately upon detecting this kind of error rather than repeatedly attempting each criterion, so you get feedback right away instead of waiting for the full review cycle to complete.

Issue: Some students were unable to save their survey responses. Matrix-type questions (where you select answers in a grid/table) were particularly affected — selections would either fail to save entirely, or certain rows would appear blank after returning to the survey.

What was fixed:

Survey responses now save reliably for all question types, including matrix questions with longer row labels.

Matrix selections that previously appeared to "reset" or disappear after leaving and returning to a survey will now persist correctly.

What users may need to do:

Students who previously had trouble saving matrix question responses may find that one or two specific rows still appear unselected. They will need to re-select those rows one time and save. After that, all selections will persist normally going forward.

New Release Alpha 0.10.7