minor docs update

azureml/medimageinsight/adapter-training.ipynb

@@ -2423,7 +2423,11 @@
    "source": [
     "## 5. Final Remarks\n",
     "\n",
-    "The adapter training approach significantly improves upon the initial zero-shot classification, particularly in terms of accuracy and the ability to differentiate between various pathologies. In this particular example we observe potential for further refinement by improving the model's specificity when it comes to classifying Atelectasis. In case if this were the case in real world, the next steps to consider for achieving better generalization to real-world scenarios would be introducing data augmentation, improving dataset diversity, end eventually performing end-to-end fine-tuning. "
+    "The adapter training approach significantly improves upon the initial zero-shot classification, particularly in terms of accuracy and the ability to differentiate between various pathologies. In this particular example we observe potential for further refinement by improving the model's specificity when it comes to classifying Atelectasis. In case if this were the case in real world, the next steps to consider for achieving better generalization to real-world scenarios would be introducing data augmentation, improving dataset diversity, end eventually performing end-to-end fine-tuning.\n",
+    "\n",
+    "## Next Steps\n",
+    "\n",
+    "Having learned about the individual steps, you can explore the [MedImageInsights Fine-tuning](./finetuning/mi2-finetuning.ipynb) notebook for an example on how to fine-tune the underlying MedImageInsights model, and for details on how to run the fine-tuning and adapter training pipelines in AzureML, using pre-built pipeline components."
    ]
   }
  ],

azureml/medimageinsight/zero-shot-classification.ipynb

@@ -501,9 +501,22 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "**Final Remark:**\n",
+    "## Final Remark:\n",
     "\n",
-    "Moving forward, we will focus on improving the model’s performance using the same set of embeddings but applying a simple adaptor to enhance classification accuracy, particularly for the more classes such as \"Support Devices.\" In the [Adapter Training](./adapter-training.ipynb), we will quantify and demonstrate the improvements in performance after integrating the adaptor model. This step will help refine the system’s ability to distinguish between closely related classes and reduce misclassification rates, setting the stage for better results in the classification of chest X-ray images."
+    "Moving forward, we will focus on improving the model's performance using the same set of embeddings but applying a simple adaptor to enhance classification accuracy, particularly for the more classes such as \"Support Devices.\" In the [Adapter Training](./adapter-training.ipynb), we will quantify and demonstrate the improvements in performance after integrating the adaptor model. This step will help refine the system's ability to distinguish between closely related classes and reduce misclassification rates, setting the stage for better results in the classification of chest X-ray images.\n",
+    "\n",
+    "## See Also\n",
+    "\n",
+    "### Continue Your Workflow with Healthcare AI Model Evaluator (HAIME)\n",
+    "\n",
+    "Now that you've established a zero-shot baseline and generated predictions with MedImageInsight, continue your workflow **Healthcare AI Model Evaluator (HAIME)** to accelerate your journey from experimentation to annotation, evaluation, and model fine-tuning. HAIME is an open-source evaluation platform designed for healthcare AI, enabling multi-disciplinary teams to build and validate models on their own data and tasks with subject matter experts in the loop.\n",
+    "\n",
+    "To explore this further, check how to [continue this workflow with HAIME](https://aka.ms/haime-zero-shot-workflow). In the guide you will learn to:\n",
+    "\n",
+    "- **Leverage your existing predictions**: Import the `zero_shot_classification_results.csv` you just generated for immediate evaluation.\n",
+    "- **Multi-reviewer evaluation**: enable _human-in-the-loop validation_ by collecting feedback from Subject Matter Experts, and combine expert evaluation with _Model reviewers_ for comprehensive assessment.\n",
+    "- **Compute transparent metrics** with support for custom metrics that match your use-case.\n",
+    "- **Bridge to improvement**: Export annotations and evaluations for downstream training and analysis."
    ]
   },
   {