# Release notes > The **UiPath.DocumentUnderstanding.ML.Activities** package is automatically installed with the **UiPath.IntelligentOCR.Activities** package to perform ML-based classification and extraction. ## Addendum: Added February 3, 2026 The **UiPath.DocumentUnderstanding.ML.Activities** package is automatically installed with the **UiPath.IntelligentOCR.Activities** package to perform ML-based classification and extraction. Starting 2026, the release notes for the Document Understanding ML activity package will be provided in the [Intelligent OCR](https://docs.uipath.com/activities/other/latest/document-understanding/release-notes-uipath-intelligent-ocr-activities#release-notes) activities section. :::note Do not install **UiPath.DocumentUnderstanding.ML.Activities** package as a separate package in you automation as the correct version will be automatically installed when you install the **UiPath.IntelligentOCR.Activities** package. ::: ## v1.36.3 **Release date: March 16, 2026** ### **Improvements** Improved performance for PDF processing, reducing digitization time for large files. ## v1.36.2 **Release date: December 16, 2025** ### What's new **Support for name-based binding and remappable IXP project resources in Maestro** We added support for referencing IXP projects by name in Maestro and introduced the ability to remap the project as a resource during deployment. ## v1.33.0 **Release date: April 16, 2025** ### What's new ### Enhanced extraction capabilities using new generative extractors To enhance generative extraction capabilities for both short and long documents that have complex layouts, you can now select the following generative extractors when configuring extractors for your document fields: * **Short Documents Simple Layout** – This is the existent Generative Extractor. * **Long Documents Complex Layout** (Preview) – Optimized for long form documents that include images, handwriting, form elements or other complex layouts, such as floating callout boxes. Examples of documents that are suitable for this extractor: insurance policies, or other similar long form documents with complex layouts. * **Short Document Complex Layout** (Preview) – Optimized for short documents that include images, handwriting, form elements or other complex layouts like floating callout boxes. For example: identity cards, or healthcare intake forms with complex layouts. ### Improvements * When using the Generative Extractor inside a Data Extraction Scope, the **Manage Prompt** button has been renamed to **Manage Field Details**. Additionally, inside the **Generative Extractor Prompt** wizard, the **Prompt** value has been renamed to **Field details**. These changes accurately describe purpose of the instructions given to the generative extractors for a specific field. * When using the Generative Classifier inside a Classify Document Scope, the **Manage Prompt** button has been renamed to **Manage Document Types**. Additionally, inside the **Generative Classifier Prompt** wizard, the **Prompt** value has been renamed to **Document Type details**. These changes accurately describe the purpose of the instructions given to the generative classifier for a specific document type. ## v1.31.2 **Release date: November 26, 2024** ### Bug fixes In the **Configure Extractors** menu, selecting **Get or refresh extractor capabilities** for an extractor trainer under the **Train Extractors Scope** activity did not open as expected. ## v1.24.1 **Release date: October 21, 2024** ### Bug fixes This version resolves the **Unknown** sections in the AI Units consumption overview dashboard that were caused by missing data. ## v1.28.8 **Release date: October 14, 2024** ### Bug fixes This version resolves the **Unknown** sections in the AI Units consumption overview dashboard that were caused by missing data. ## v1.31.1 **Release date: October 3, 2024** :::note This package version is compatible with UIAutomation.Activities versions that are 24.10.0 or later. ::: ### What's new ### Support for activities from an on-premises setup You can now use the [Generative Classifier](https://docs.uipath.com/activities/other/latest/document-understanding/document-understanding-ml-generative-classifier#generative-classifier) and [Generative Extractor](https://docs.uipath.com/activities/other/latest/document-understanding/generative-extractor#generative-extractor) activities within a [Classify Document Scope](https://docs.uipath.com/activities/other/latest/document-understanding/classify-document-scope) and [Data Extraction Scope](https://docs.uipath.com/activities/other/latest/document-understanding/data-extraction-scope), even if the robot is connected to a local Orchestrator. We've added the **RuntimeTenantURL** and **RuntimeCredentialsAsset** properties to the Generative Classifier and Extractor activities. With these properties, you can now directly use credentials from external applications, stored in Orchestrator, to access Document Understanding resources at runtime. To achieve this, ensure that your selected tenant has Document Understanding enabled and AI Units allocated. ### Improvements Increased prompt size from 500 to 1000 characters per question for enhanced clarity in your instructions. ### Bug fixes The ML extractor returned empty values, causing validation tasks to fail due to null value properties. ## v1.28.7 **Release date: 13 August 2024** We've upgraded some internal dependencies for enhanced performance. ## v1.28.6 **Release date: 31 July 2024** ### Bug fixes When using the Generative Classifier and Generative Extractor activities, inputting a prompt that started or ended with whitespaces led to a "KeyNotFoundException" error with the "The given key was not present in the dictionary" message. ## v1.28.5 **Release date: 12 July 2024** ### Bug fixes Fixed the issue causing document classification to fail with the "An invalid request URI was provided" error. ## v1.28.4 **Release date: 20 June 2024** We are constantly working to improve your UiPath Document Understanding experience. Even though there are no major significant changes with this release, we made sure to bring minor improvements and accessibility fixes to our product. ## v1.28.3 **Release date: 5 June 2024** We've improved product stability by revising certain dependencies. ## v1.28.2 **Release date: 27 May 2024** ### Improvements Increased prompt size from 500 to 1000 characters per question for enhanced clarity in your instructions. Also, if you reach the prompt size limit of 1000 characters per question, you will receive a "Limit exceeded" error. ### Bug fixes The ML extractor returned empty values, causing validation tasks to fail due to null value properties. ### Known issues If the Generative Extractor or Classifier triggers a `Content Filtered` exception, the activity does not generate any results, as if the content was missing. The following warning message will show in the robot logs: `GPT refused to handle the request because of content filtering policy. Returning empty result.` This message is also displayed in Studio when an automation is initiated from there. ## v1.28.1 **Release date: 29 April 2024** ### New features The Machine Learning Extractor can now extract vertical text, significantly enhancing the extraction capabilities. ## v1.24.0 **Release date: 1 November 2023** ### Generative Features General Availability We are excited to announce that the **UiPath.DocumentUnderstanding.ML.Activities** generative features are now in general availability. This includes the following: * Generative Classifier activity * Generative Extractor activity :::tip You can use the new generative features with the **UiPath.IntelligentOCR.Activities** framework. ::: ### New Features and Improvements * A new property, **Output Folder**, is available for the Machine Learning Classifier Trainer activity. This property allows you to save files locally. * Support for multi-page fields is now available. This feature is useful when an Address, for example, has the street on one page and the state and zip code on the following page. Due to a known issue, table rows are currently not working in this case. Follow our release notes for updates in the future. ## v1.21.2 **Release date: 7 June 2023** We are constantly working to improve your UiPath Document Understanding experience. Even though there are no major significant changes with this release, we made sure to bring minor improvements and accessibility fixes to our product. ## v1.21.1 **Release date: 26 April 2023** ### Deprecation Timeline We recommend that you regularly check the [deprecation timeline](https://docs.uipath.com/overview/other/latest/overview/deprecation-timeline#activities-upcoming-removals) for any updates regarding features that will be deprecated and removed. ## v1.17.1 **Release date: 27 March 2023** ### New features & Improvements We are constantly working to improve the UiPath.DocumentUnderstanding.ML.Activities package. Even though there are no major significant changes with this patch release, we made sure to bring minor improvements and accessibility fixes to our product. ## v1.18.0 **Release date: 15 December 2022** ### New features & Improvements * The UiPath Studio user interface is now available in Traditional Chinese. * You can now benefit from the **API Key** field being pre-populated for the following activities included in the UiPath.DocumentUnderstanding.ML.Activities package: **Machine Learning Classifier** and **Machine Learning Extractor**. ## v1.17.0 **Release date: 24 October 2022** ### New features & Improvements * The **Machine Learning Extractor Trainer** activity can now support multivalued fields. * The **UseServerSideOCR** option is scheduled to be deprecated on December 2022. We recommend using the default behavior. More details about the deprecation can be found [here](https://docs.uipath.com/overview-guide/docs/deprecation-timeline#activities). * The Document Understanding Process Studio template has been upgraded to a new version. The **UiPath.DocumentUnderstanding.ML.Activities** package is a dependency for this template. ### Bug Fixes * Fixed a bug that was causing extraction errors when Digitizer was used by upgrading the PDF library and using hybrid OCR features. * Fixed a bug that caused inconsistent input to be sent to ML Extractor when both image and DOM are required. * The ProxySettings were not used in the `GetCapabilities` call received from Machine Learning Extractor. The bug is now fixed and works as expected. ## v1.13.2 **Release Date: 9 May 2022** ### Improvements * The **UiPath.DocumentUnderstanding.ML.Activities** package has been upgraded to .NET5 portable, allowing you to run them on Linux robots. * The **Machine Learning Extractor Trainer** and the **Machine Learning Classifier Trainer** activities have received new parameters grouped under the name of **Public Datasets**, allowing you to use public datasets instead of private ones. * The **Machine Learning Extractor** activity has been updated and now the extraction algorithm can also be used from **Forms AI**, not only from the **ML Models** list. * The **Machine Learning Extractor** activity can now be used with a public endpoint in airgapped scenarios. ### Bug Fixes Fixed a bug occurring on the **Machine Learning Extractor Trainer** activity that was exporting values even if the field was not mapped in the **Configure Trainer** wizard. Now, everything works as expected. ## v1.9.2 **Release Date: 3 August 2023** ### Bug Fix In the Document Understanding ML Activities Package, some of the fields that were required were not marked accordingly. Now all mandatory fields are visually marked, as expected. ## v1.9.1 **Release Date: 5 October 2021** ### New Features and Improvements * This release brings as an improvement the telemetry client update to version 1.5.3. * The **UiPath.DocumentUnderstanding.ML.Activities** package has been upgraded to .NET5. While both .NET versions continue to be supported, the .NET5 projects can only work on 64-bit architectures. * Added the **Dataset** and **Project** parameters to the **Machine Learning Extractor Trainer** activity which allow you to select where to upload your training data in your AI Center tenant. As a result, the **Endpoint** and **MLSkill** parameters were removed from the activity. * Added the **Endpoint** parameter to the **Machine Learning Classifier** activity which provides the ability to use the activity with public ML Skills. * The **Machine Learning Extractor** can now be integrated with Forms AI. The only requirement for this to happen is to make sure that the **UseServerSideOCR** option is disabled. ## v1.7.0 **Release Date: 29 March 2021** ### New Features and Improvements * Released the [Machine Learning Classifier](https://docs.uipath.com/activities/other/latest/document-understanding/machine-learning-classifier#machine-learning-classifier) and [Machine Learning Classifier Trainer](https://docs.uipath.com/activities/other/latest/document-understanding/machine-learning-classifier-trainer#machine-learning-classifier-trainer) activities as part of the Machine Learning Document Classification functionality which helps you classify documents using a custom trained ML model. **Machine Learning Classifier** could prove to be very useful particularly in scenarios with high diversity in document sets. To train the classifier and improve its results with time with the aid of human validation, you can use the *sister* activity, **Machine Learning Classifier Trainer**. * Improved processing of PDF files. ## v1.5.2 **Release Date: 11 November 2020** ### Bug Fixes * In case of failure response when requesting the ML Skills list from AI Fabric, automatic retry is now available for 5XX HTTP status codes as well. * In case of network failure instead of the retry operation being blocked, the **Machine Learning Extractor** now retries to connect to AI Fabric. ## v1.5.1 **Release Date: 20 October 2020** ### Improvements More detailed error logging for **Machine Learning Extractor**. ### Bug Fixes * Added retry mechanism for ML Skills requests. * Fixed an issue where not all ML Skills were usable by the **Machine Learning Extractor** activity. ## v1.5.0 **Release Date: 5 October 2020** ### New Features and Improvements * Released the new **Machine Learning Extractor Trainer** activity can prepare data for ML model re-training based on human validation results. * Added the **Get or refresh extractor capabilities** functionality to **Machine Learning Extractor Trainer** that can be used to easily map your taxonomy fields with the available extractor fields. * A new parameter has been included in the **Machine Learning Extractor** activity, named **Timeout (milliseconds)**. The parameter can be used for specifying the amount of time to wait for a response from the server before an error is thrown. * Changed the tooltip text on **UseServerSideOCR** property for **Machine Learning Extractor** to indicate it is incompatible with **Machine Learning Extractor Trainer**. ## v1.2.2 **Release Date: 24 August 2020** ### Bug Fixes * Fixed an issue that in some cases was returning a `407ProxyAuthenticationRequired` error message for Kerberos or NTLM authentication requests. This applies to **Machine Learning Extractor**. * Fixed an issue that was causing the **Get Capabilities** functionality of **Machine Learning Extractor** not to work if a certain endpoint was provided. * Fixed an issue that was causing the **Machine Learning Extractor** to throw an error when no robot is connected. ## v1.2.1 **Release Date: 5 June 2020** ### New features and improvements This release brings a new option available for the **Machine Learning Extractor** activity, named MLSkill, that can be used for consuming models stored in AI Fabric. ## v1.1.0 **Release Date: 4 May 2020** ### New Features and Improvements A new option has been added to the **Machine Learning Extractor** named `UseServerSideOCR`. This option allows you to use the OCR results received from digitization. The **Machine Learning Extractor** now declares its internal taxonomy, allowing you to easily map the fields it can extract to the fields you have defined in your taxonomy, in the **Configure Extractors** wizard of the **Data Extraction Scope**. ## v1.0.0 **Release Date: 11 February 2020** ### New Features and Improvements The **UiPath.DocumentUnderstanding.ML.Activities** package facilitates the automation of information extraction from receipts and invoices while using AI methods.