Benchmarks can now be fuzzy. For example, a project manager may mark more than one correct answer to a radio or dropdown classification.. Read more in our docs

In projects where benchmarking has been activated, the Tasks tab now shows each benchmark’s average score and the number of users who have annotated each benchmark.

In the labeling editor, pressing Ctrl + A will now select all objects in the asset (or all objects for the current frame in multi-frame assets).

In the Hold and Start stages, users can now type to search names of batches to forward to the next stage.

Task attachments can now contain arbitrary HTML.. Read more in our docs

Hiding or revealing objects with the keyboard shortcut “H” will no longer delete an annotation that is currently being drawn with the mouse cursor.

Answers to multiple classification questions no longer have a line running between the classification title and its options, improving readability in projects with large numbers of multiple classification answers.

The code powering the Segmentation tool has been substantially rewritten, leading to possible performance improvements in some scenarios and simplifying further development on the tool.

Idle duration and blur duration times can now be optionally included in the export.

The way total, idle, and blur durations are displayed in the Task Info panel has been improved.

Frames per second information in the Task Info view is now limited to two decimals.

The information panel regarding the transition to Export V3 has been removed from the Export tab.

When deleting assets from the Settings tab, files related to assets uploaded with drag-and-drop will also be purged from iMerit’s AWS private S3 storage. Read more in our docs.

OCR text can now be imported in the pre-labeling step for bounding box tools which have the Targeted OCR toggle activated. Read more in our docs.