fixes

Author: lgayhardt

articles/ai-studio/concepts/evaluation-approach-gen-ai.md

@@ -80,10 +80,11 @@ Cheat sheet:
 
 | Purpose |  Process | Parameters |
 | -----| -----| ----|
-| What are you evaluating for? | Identify or build relevant evaluators | - [Quality and performance](./evaluation-metrics-built-in.md?tabs=warning#generation-quality-metrics) ( [Quality and performance sample notebook](https://github.com/Azure-Samples/rag-data-openai-python-promptflow/blob/main/src/evaluation/evaluate.py)) </br> - [Safety and Security](./evaluation-metrics-built-in.md?tabs=warning#risk-and-safety-metrics)) ([Safety and Security sample notebook]((https://github.com/Azure-Samples/rag-data-openai-python-promptflow/blob/main/src/evaluation/evaluatesafetyrisks.py)))  </br> [Custom](../how-to/develop/evaluate-sdk.md#custom-evaluators) ([Custom sample notebook](https://github.com/Azure-Samples/rag-data-openai-python-promptflow/blob/main/src/evaluation/evaluate.py))]  |
-| What data should you use?  | Upload or generate relevant dataset | [Generic simulator for measuring Quality and Performance](./concept-synthetic-data.md) ( Generic simulator sample notebook|(https://github.com/Azure/azureml-examples/blob/main/sdk/python/foundation-models/system/finetune/Llama-notebooks/datagen/synthetic-data-generation.ipynb)] </br> - Adversarial simulator for measuring Safety and Security [[Adversarial simulator Docs](../how-to/develop/simulator-interaction-data.md), [Adversarial simulator sample notebook](https://github.com/Azure-Samples/rag-data-openai-python-promptflow/blob/main/src/evaluation/simulate_and_evaluate_online_endpoint.ipynb)]  |
-| What resources should conduct the evaluation? | Run evaluation | - Local run  <br> </br>  - Remote cloud run |
+| What are you evaluating for? | Identify or build relevant evaluators | - [Quality and performance](./evaluation-metrics-built-in.md?tabs=warning#generation-quality-metrics) ( [Quality and performance sample notebook](https://github.com/Azure-Samples/rag-data-openai-python-promptflow/blob/main/src/evaluation/evaluate.py)) </br> - [Safety and Security](./evaluation-metrics-built-in.md?tabs=warning#risk-and-safety-metrics)) ([Safety and Security sample notebook]((https://github.com/Azure-Samples/rag-data-openai-python-promptflow/blob/main/src/evaluation/evaluatesafetyrisks.py)))  </br> [Custom](../how-to/develop/evaluate-sdk.md#custom-evaluators) ([Custom sample notebook](https://github.com/Azure-Samples/rag-data-openai-python-promptflow/blob/main/src/evaluation/evaluate.py))] |
+| What data should you use?  | Upload or generate relevant dataset | [Generic simulator for measuring Quality and Performance](./concept-synthetic-data.md) ( [Generic simulator sample notebook|(https://github.com/Azure/azureml-examples/blob/main/sdk/python/foundation-models/system/finetune/Llama-notebooks/datagen/synthetic-data-generation.ipynb)] </br> - Adversarial simulator for measuring Safety and Security [Adversarial simulator Docs](../how-to/develop/simulator-interaction-data.md) (Adversarial simulator sample notebook](https://github.com/Azure-Samples/rag-data-openai-python-promptflow/blob/main/src/evaluation/simulate_and_evaluate_online_endpoint.ipynb)] )  |
+| What resources should conduct the evaluation? | Run evaluation | - Local run  </br>  - Remote cloud run |
 | How did my model/app perform? | Analyze results  | [View aggregate scores, view details, score details, compare eval runs](..//how-to/evaluate-results.md)] |
+| How can I improve? | Make changes to model, app, or evaluators  - If evaluation results did not align to human feedback, adjust your evaluator. </br> - If evaluation results aligned to human feedback but did not meet quality/safety thresholds, apply targeted mitigations. |
 
 ## Related content