6.69.0
July 04, 2024
What's new 💡
NLP
Added supported file types information in Step 1 of the Project Creation Wizard for Bounding Box Labeling: YOLO, LabelMe, and Pascal VOC.
Adjust the dialog and snack bar messages for creating and editing Bounding Boxes to ensure clearer information.
Custom Report Builder: A new group table called Progress. It can be used to track your labeling work which contains these metrics: Total Documents Completed, Total Documents Answered, Total Rows Answered, and Total Sentences Labeled.
Improve the appearance of the delete button on the Label Management page to avoid cropping issues.
Multiple Target Texts and Target question for ML Assisted - LLM Labs. Users can now provide more context to their LLM application in LLM Labs and receive multiple predictions in JSON format. These predictions can be sent to the NLP Platform using ML Assisted.
Rename the question-answering extension for Row Labeling project. Previously, it was called
Document labeling
extension.Switching from flat comments to threaded comments in the editor for Span Labeling, with replies automatically expanded by default to enhance the user experience.
LLM Labs
Added a new automated evaluation feature.
Added a new feature to let users choose from ground truth when selecting datasets in the automated evaluation creation wizard.
Added a new error message for when there is an error creating a vector store.
Added a new copy button feature in LLM Playground when generating code as a completion result.
Added OpenAI embedding models as direct access models.
Added GPT-4 from OpenAI provider as a new Direct Access LLMs.
Equalized the display of vector store URL documents based on their paths.
Improved checkbox consistency in the LLM playground after page refresh.
Improved the file preview for .docx files in Vector Store to prevent cropping.
Improved the behavior when users click the "retry failed vector store" button; the user page will stay the same and not be redirected.
Users will be redirected to the correct module they are currently working on.
Bug fixes 🐞
NLP
A more accurate calculation of Total Time Spent by excluding the idle time of each user.
Fixed an issue in OCR projects (Span Labeling with documents as media files) where documents were displayed cropped when resized.
Fixed an issue where the context menu was still shown for rejected labels even when the 'Show Rejected Labels in Review Mode' option was enabled.
Fixed the display in the reply box in comments to ensure resolved comments are displayed correctly and do not appear cropped.
Fixed an issue in Step 3 of the Project Creation Wizard where the tooltip incorrectly listed file types other than .CSV for LLM Ranking and LLM Evaluation projects.
Fixed the answers not updated in Document Labeling extension on Predictive Labeling.
Improving ML Assisted with LLM Labs Error Message.
LLM Labs
Fixed an issue where the evaluation list table displayed no results when a manual evaluation record was present.
Fixed an issue where LLM Evaluation returned a blank response after undeploying an application.
Last updated