Skip to content

Commit d30a50e

Browse files
committed
add iteration
1 parent 16af1a3 commit d30a50e

File tree

1 file changed

+9
-0
lines changed

1 file changed

+9
-0
lines changed

articles/ai-studio/tutorials/copilot-sdk-evaluate.md

Lines changed: 9 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -152,6 +152,15 @@ You can also look at the individual rows and see metric scores per row, and view
152152
153153
For more information about evaluation results in AI Studio, see [How to view evaluation results in AI Studio](../how-to/evaluate-results.md).
154154
155+
## Iterate and improve
156+
157+
Notice that the responses are not well grounded. In many cases, the model replies with a question rather than an answer. This is a result of the prompt template instructions.
158+
159+
* In your **assets/grounded_chat.prompty** file, find the sentence "If the question is related to outdoor/camping gear and clothing but vague, ask for clarifying questions instead of referencing documents." * Change the sentence to "If the question is related to outdoor/camping gear and clothing but vague, try to answer based on the reference documents, then ask for clarifying questions."
160+
* Save the file and re-run the evaluation script.
161+
162+
Try other modifications to the prompt template, or try different models, to see how the changes affect the evaluation results.
163+
155164
## Clean up resources
156165
157166
To avoid incurring unnecessary Azure costs, you should delete the resources you created in this tutorial if they're no longer needed. To manage resources, you can use the [Azure portal](https://portal.azure.com?azure-portal=true).

0 commit comments

Comments
 (0)