# 6.90.0

November 28, 2024

## **What's new** 💡

### **Data Studio**

* Bounding Box Labeling — Users can now [hide bounding boxes by label class](https://github.com/vulcan-ai/gitbook/blob/datasaur/compatibility-and-updates/nlp-projects/lets-get-labeling/bounding-box-labeling.md#how-to-hide-bounding-boxes), making it easier to edit overlapping annotations during labeling tasks.
* Improved the reset functionality to ensures labels, timestamps, and bounding boxes are immediately cleared without requiring a page refresh across Span, Audio, and OCR Labeling projects.
* List of File Extension — Introducing a new labeler status indicator to help users easily identify files assigned to them, enhancing clarity in the review process.
* [Self-Assignment](https://github.com/vulcan-ai/gitbook/blob/datasaur/compatibility-and-updates/workspace-management/project-management/self-assignment/README.md) — Introducing a new way for labelers to select and assign documents to themselves from available projects with incomplete tasks.

### **LLM Labs**

* Auto-generate instruction — Improved the scroll behavior of the action button.
* Deployment — Improved the deployed application to return the retrieved chunk.
* Deployment — Removed the application name from the deployed application endpoint.
* Direct access LLMs — Integrated a new Diffbot model into our Direct access LLMs.
* Evaluation — Improved the failed evaluation result to display as dashed lines in the evaluation details.
* Knowledge base — Improved our chunking mechanism to not isolate chunks by content pages.

## **Bug fixes** 🐞

### **Data Studio**

* Evaluation Metrics — Fixed an issue where evaluation metrics were not displayed for single-question Row Labeling projects with only two options.

### **LLM Labs**

* Fixed an issue where the custom model was not returning a stopReason response.
