Merge pull request #28950 from MicrosoftDocs/lauragra-issue968

Update minimum number of files for training.

Author: Ruchika-mittal01

copilot/copilot-tuning-doc-generation.md

@@ -5,7 +5,7 @@ ms.author: jasonjoh
 manager: calvind
 ms.audience: ITPro
 ms.reviewer: jwolk
-ms.date: 06/17/2025
+ms.date: 07/17/2025
 ms.topic: how-to
 ms.localizationpriority: medium
 description: Learn how to use Copilot Tuning to build an AI model for document generation based on organizational knowledge.
@@ -33,9 +33,9 @@ Some example use cases include:
 
 - You must have permission to use Copilot Tuning in Copilot Studio. <!-- TODO: Link to permission doc here if it exists -->
 - A collection of original documents and corresponding final draft documents that are stored in SharePoint.
-- A collection of change logs or specifications stored in SharePoint.
+- A collection of changelogs or specifications stored in SharePoint.
 - A structured version of required changes to provide in the supplementary field within Copilot Tuning.
-- A minimum of 20 well-aligned pairs of reference documents to target pairs that reflect a representative range of changes you expect the system to handle.
+- More than 20 well-aligned pairs of reference documents to target pairs that reflect a representative range of changes you expect the system to handle.
 
 > [!IMPORTANT]
 > Document generation supports working with the following file formats: .doc, .docx, .html, .md, or .pdf. Copilot Tuning only uses information found in text. Copilot Tuning document generation doesn't use information in images, tables, or unstructured web content in your documents.
@@ -55,7 +55,7 @@ The following are the high-level steps to configure a custom document generation
 
 ### Prepare a mapping file
 
-Your knowledge source should have at least 20 example pairs of original files and corresponding final (draft) files. In this step, you prepare a CSV file that provides at least 20 examples of original files to final (draft) documents. Copilot Tuning uses these examples to fine-tune the generation logic, helping the model learn how your organization typically edits or adapts documents.
+Your knowledge source should have more than 20 example pairs of original files and corresponding final (draft) files. In this step, you prepare a CSV file that provides more than 20 examples of original files to final (draft) documents. Copilot Tuning uses these examples to fine-tune the generation logic, helping the model learn how your organization typically edits or adapts documents.
 
 Create a file named **mapping.csv** and store it in the root directory of your knowledge source. This file should have two columns:
 

copilot/copilot-tuning-expert-qa.md

@@ -4,7 +4,7 @@ author: danielabom
 ms.author: danielabo
 manager: calvind
 ms.reviewer: jwolk
-ms.date: 06/17/2025
+ms.date: 07/16/2025
 ms.topic: how-to
 ms.localizationpriority: medium
 description: Learn how to use Copilot Tuning to build an AI model for expert question & answer (Q&A).
@@ -29,19 +29,17 @@ You can use the model maker to fine-tune a model that can complete the following
 
 ## Limitations
 
-The model supports various document formats, but there are specific limitations to consider when using your content.
+The model supports various document formats, with the following specific limitations:
 
 - Content must be stored in SharePoint and be in supported formats (.docx, .pdf, .aspx); elements like embedded images or tables aren't supported.
 - Not intended for general productivity or web-wide knowledge queries; it's limited to tenant-specific content and not suitable for tasks like managing meetings or browsing general internet data.
 - Depending on the snapshot time of training data, newer content must be enriched via Search.
 
-While there's no minimum document count, better results are achieved with larger content sets; at least 20 documents are recommended for training.
-
 ## Prerequisites
 
 Before you start, make sure that you have the following prerequisites in place:
 
-1. You must have domain-specific content or documentation, such as legal playbooks, HR guidelines, technical documentation, policy manuals, or departmental procedures, that the model can use to answer questions.
+1. You must have domain-specific content or documentation, such as legal playbooks, HR guidelines, technical documentation, policy manuals, or departmental procedures, that the model can use to answer questions. The content set that you use to train the model must consist of more than 20 files.
 2. Configure the model agent with a Microsoft Entra ID security group or distribution list and create your own Entra ID groups to be added to the model.
 3. Identify where your content is stored in SharePoint.
 

copilot/copilot-tuning-knowledge-selection.md

@@ -5,7 +5,7 @@ author: msplants
 ms.author: jwolk
 manager: calvind
 ms.topic: concept-article
-ms.date: 06/17/2025
+ms.date: 07/17/2025
 ms.localizationpriority: medium
 ---
 
@@ -25,7 +25,7 @@ Selecting knowledge is the first and most critical step in Copilot Tuning. You s
 
 - **Supported file formats and content:** All Copilot Tuning tasks support common text-based document formats. You can use Word documents (.doc, .docx), HTML files (.html, .aspx), Markdown files (.md), or PDF files processed by Optical Character Recognition (OCR) as source materials. You can also include Excel documents (.xls, .xlsx) for expert Q&A. The tuning process ingests the text content from these files. It doesn't learn from images, embedded tables, or other non-text elements in the documents. Ensure that the important information in your training documents is in textual form. For example, if a filePDF contains a chart, include a textual explanation of that chart's insights in the document.
 
-- **Number of documents:** You must provide at least 20 samples (documents for Q&A and summarization; input-output pairs for document generation) to Copilot Tuning. Usually hundreds or thousands of samples is ideal, and you can provide a maximum of 10k. The quality of samples is more important than raw quantity. We highly suggest you focus you data preparation time on finding as many high quality samples that are well-aligned with what you expect your fine-tuned model to do.
+- **Number of documents:** You must provide more than 20 samples (documents for Q&A and summarization; input-output pairs for document generation) to Copilot Tuning. Usually hundreds or thousands of samples is ideal, and you can provide a maximum of 10k. The quality of samples is more important than raw quantity. We highly suggest you focus you data preparation time on finding as many high quality samples that are well-aligned with what you expect your fine-tuned model to do.
 
 - **Model instructions:** During the model configuration process, Copilot Tuning asks the model maker to provide answers to a series of model instructions to guide the system on how to use the knowledge you selected. Each task type has its own questions about the selected knowledge source. Prepare clear, structured answers to each question. Expert Q&A requires a description of the data in the knowledge source and how it's organized. Document generation requires you to specify how the original input, changes, and output draft document are referred to in your organization. Summarization requires you to specify how to refer to the summaries. It's important that this information is clear and accurately represents your data in order for the system to be most effective.
 

copilot/copilot-tuning-troubleshooting-mac.md

@@ -4,7 +4,7 @@ f1.keywords:
 ms.author: emrek
 author: emrekiciman
 manager: calvind
-ms.date: 07/08/2025
+ms.date: 07/17/2025
 audience: Admin
 ms.topic: troubleshooting
 ms.service: microsoft-365-copilot