# 6.138.0

November 13, 2025

## **What's New? 💡**

### **Data Studio**

* Span Labeling Editor — Introduced a collapsible toolbar and collapsed extension sidebar by default, providing a cleaner interface. The toolbar automatically appears when applying filters or sorting, and hides when not in use, giving users more vertical space for labeling.
* [Pre-labeled Answers in Row Labeling](https://docs.datasaur.ai/data-studio-projects/pre-labeled-project#pre-labeled-propagation) — Added a new setting that requires labelers to review and approve pre-labeled answers before they are counted as their work.
* Search extension in Span Labeling — Add the capability to [delete or reject labels](https://docs.datasaur.ai/advanced/extensions/search#delete-search-result-labels), enabling bulk delete directly from search results.

### **LLM Labs**

* [Fine-tuning](https://docs.datasaur.ai/llm-projects/models/fine-tuning) — Added a new capability to show relevant metrics during the fine-tuning process, providing better visibility into model performance.
* [Dataset](https://docs.datasaur.ai/llm-projects/dataset) — Implemented a helpful tooltip on the Upload Dataset button to provide users with additional guidance and information.

## **Bug Fixes 🐞**

### **Data Studio**

* Fixed an issue in Bounding Box Labeling where labeling or completing documents could trigger out-of-memory errors.
* Fixed an issue in Span Labeling where pre-labeled labels did not appear in the editor after project creation.
* Fixed an issue in Row Labeling where rejecting a labeler-added line didn’t also reject the related answers.

### **LLM Labs**

* Fixed an issue when setting JSON output structure in model hyperparameters while integrating models with knowledge bases.
* Fixed an issue where "Unexpected error in evaluation!" message was appearing for all rows during RAGAS evaluation.
* Fixed a document parsing exception in deployed application logging.