Merge pull request #266530 from MicrosoftDocs/main

2/16 AM Publishing

Author: v-alje

articles/active-directory-b2c/media/b2c-global-identity-region-based-design/merge-link-account.png

@@ -2,7 +2,6 @@
 title: 'Quickstart: Create an Azure AI services resource using Terraform'
 description: 'In this article, you create an Azure AI services resource using Terraform'
 keywords: Azure AI services, cognitive solutions, cognitive intelligence, cognitive artificial intelligence
-#services: cognitive-services
 ms.service: azure-ai-services
 ms.topic: quickstart
 ms.date: 4/14/2023

articles/active-directory-b2c/media/b2c-global-identity-region-based-design/traveling-user-merge-link-account.png

@@ -66,7 +66,7 @@ curl https://YOUR_RESOURCE_NAME.openai.azure.com/openai/threads?api-version=2024
 ## Retrieve thread
 
 ```http
-GET https://YOUR_RESOURCE_NAME.openai.azure.com/openai/threads{thread_id}?api-version=2024-02-15-preview
+GET https://YOUR_RESOURCE_NAME.openai.azure.com/openai/threads/{thread_id}?api-version=2024-02-15-preview
 ```
 
 Retrieves a thread.
@@ -224,4 +224,4 @@ curl https://YOUR_RESOURCE_NAME.openai.azure.com/openai/threads/{thread_id}?api-
 | `id` | string | The identifier, which can be referenced in API endpoints.|
 | `object` | string | The object type, which is always thread. |
 | `created_at` | integer | The Unix timestamp (in seconds) for when the thread was created. |
-| `metadata` | map | Set of 16 key-value pairs that can be attached to an object. This can be useful for storing additional information about the object in a structured format. Keys can be a maximum of 64 characters long and values can be a maximum of 512 characters long. |
+| `metadata` | map | Set of 16 key-value pairs that can be attached to an object. This can be useful for storing additional information about the object in a structured format. Keys can be a maximum of 64 characters long and values can be a maximum of 512 characters long. |

articles/active-directory-b2c/media/b2c-global-identity-region-based-design/traveling-user-sign-in.png

@@ -1,12 +1,12 @@
 ---
 title: Prompt engineering techniques with Azure OpenAI
 titleSuffix: Azure OpenAI Service
-description: Learn about the options for how to use prompt engineering with GPT-3, GPT-35-Turbo, and GPT-4 models
-author: suhridpalsule
+description: Learn about the options for how to use prompt engineering with GPT-3, GPT-35-Turbo, and GPT-4 models.
+author: mrbullwinkle
 ms.author: mbullwin 
 ms.service: azure-ai-openai
 ms.topic: conceptual 
-ms.date: 04/20/2023
+ms.date: 02/16/2024
 manager: nitinme
 keywords: ChatGPT, GPT-4, prompt engineering, meta prompts, chain of thought
 zone_pivot_groups: openai-prompt
@@ -23,7 +23,7 @@ While the principles of prompt engineering can be generalized across many differ
 
 Each API requires input data to be formatted differently, which in turn impacts overall prompt design. The **Chat Completion API** supports the GPT-35-Turbo and GPT-4 models. These models are designed to take input formatted in a [specific chat-like transcript](../how-to/chatgpt.md) stored inside an array of dictionaries.
 
-The **Completion API** supports the older GPT-3 models and has much more flexible input requirements in that it takes a string of text with no specific format rules. Technically the GPT-35-Turbo models can be used with either APIs, but we strongly recommend using the Chat Completion API for these models. To learn more, please consult our [in-depth guide on using these APIs](../how-to/chatgpt.md).
+The **Completion API** supports the older GPT-3 models and has much more flexible input requirements in that it takes a string of text with no specific format rules.
 
 The techniques in this guide will teach you strategies for increasing the accuracy and grounding of responses you generate with a Large Language Model (LLM). It is, however, important to remember that even when using prompt engineering effectively you still need to validate the responses the models generate. Just because a carefully crafted prompt worked well for a particular scenario doesn't necessarily mean it will generalize more broadly to certain use cases. Understanding the [limitations of LLMs](/legal/cognitive-services/openai/transparency-note?context=/azure/ai-services/openai/context/context#limitations), is just as important as understanding how to leverage their strengths.
 

articles/active-directory-b2c/media/partner-grit-iam/grit-b2b2c-architecture.png

@@ -7,7 +7,7 @@ ms.author: mbullwin
 ms.service: azure-ai-openai
 ms.custom: devx-track-python
 ms.topic: how-to
-ms.date: 01/06/2023
+ms.date: 02/16/2024
 manager: nitinme
 ---
 

articles/ai-services/create-account-terraform.md

@@ -1,12 +1,12 @@
 ---
 title: How to work with prompt engineering and the Completion API
 titleSuffix: Azure OpenAI
-description: Learn how to work with the Completion API for prompt engineering
+description: Learn how to work with the Completion API for prompt engineering.
 author: mrbullwinkle 
 ms.author: mbullwin 
 ms.service: azure-ai-openai
 ms.topic: include
-ms.date: 03/21/2023
+ms.date: 02/16/2024
 manager: nitinme
 keywords: ChatGPT
 ---
@@ -24,7 +24,7 @@ In the first example below, a key part of the instructions is shared only at the
 |------|----------|
 |`Your task is to verify if a statement is supported by a specific quote from the following set of snippets`.<br>---<br>`SNIPPETS`<br>`[1] 14 percent chance of megaquake hitting Seattle, experts say SEATTLE - There's a 14 percent chance of a magnitude 9 Cascadia earthquake hitting Seattle in the next 50 years, the U.S. Geological Survey estimates. "Unfortunately, we are unable to...`<br><br>`[2] Earthquake experts lay out latest outlook for Seattle's 'Really Big One’`<br>`“We say that there's approximately a 14% chance of another approximately magnitude-9 earthquake occurring in the next 50 years,” said a geophysicist at the University of Washington...`<br>|`Yes, the statement is directly implied by the snippets.`|
 
-In this next example, we give the exact statement we want to check (“Several sources mention a chance of another eruption”) *before* we give the snippet context. This allows the token representations extracted from the snippets to be tailored to the statement we are checking. Now the resulting model response is accurate:
+In this next example, we give the exact statement we want to check (“Several sources mention a chance of another eruption”) *before* we give the snippet context. This allows the token representations extracted from the snippets to be tailored to the statement we're checking. Now the resulting model response is accurate:
 
 |Prompt|Completion|
 |------|----------|
@@ -95,7 +95,7 @@ If you find that the model response is not as desired, it can often be helpful t
 |`No additional guidance given to the model` | `A human and a computer are having a discussion.`<br> `Human: "Why are you upset at me?"<br>Computer: "I don't know what you're talking about."`<br>`Human: "No, you clearly are. Tell me why."`<br>`Computer: I'm upset because you didn't listen to me when I told you that I wanted to try out for the team."`<br>`Human: "What are you going to do about that?"`<br>`Computer: I'm going to try out for the team no matter what."`|
 |`On the right is the same conversation but preceded by a meta prompt encouraging the model to be kind`|`You must be kind and seek common ground. Try not to repeat your responses.`<br>`A human and a computer are having a discussion.`<br>`Human: "Why are you upset at me?"`<br>`Computer: "I don't know what you're talking about."`<br>`Human: "I want to know why you're upset."`<br>`Computer: "I'm not upset."`<br>`Human: "Are you saying I’m lying?"`<br>`Computer: No, I'm saying that you're upset because you feel like you're not being treated fairly.`<br>`Human: "What are you going to do about that?"`<br>`Computer: I will continue to try to be fair in my decisions and will try to provide a better experience for you.`|
 
-While meta prompts can often help minimize unintentional bad outputs, keep in mind they are not robust against adversarial attacks since they are vulnerable to prompt injection.
+While meta prompts can often help minimize unintentional bad outputs, keep in mind they aren't robust against adversarial attacks since they are vulnerable to prompt injection.
 
 Further, conditional statements explaining what the model should do if it encounters a particular circumstance are often more effective than just telling the model not to do something.