You can find the Triage Large Language Model (LLM) Analytics under Triage > Model > Tickets. This feature provides valuable insights into your model's performance, including the number of tickets Triage predicted, ticket status, correct predictions, overall accuracy of Triage's label predictions, and more. By reviewing these metrics, you can easily assess how well your Triage LLM is performing and identify ways to enhance its effectiveness.
Key metrics
| Name | Description |
| Predicted Tickets | Tickets where Triage successfully predicted and assigned a label. |
| Tickets Status | This shows the latest ticket status (not the status of the tickets in the past). |
| Finalized Tickets | Predicted tickets that have been closed or resolved by a human agent, finalizing the ticket status. |
|
Correct Predictions
|
Tickets where Triage's predicted label was accurate, meaning the human agent did not override it, and the ticket is now closed or resolved. |
| Accuracy | The percentage of correct predictions out of all finalized tickets. Calculated as: (Correct Predictions / Finalized Tickets) * 100 |
Table metrics
| Name | Description |
| Predicted Time | The local time when Triage predicted the label and successfully wrote it on the ticket. |
| Tickets ID | Ticket ID of the connected Help Desk |
| Ticket Title | Title of the ticket. N/A if it's missing value on the Help Desk at the time of prediction. |
|
Ticket Body |
Content of the ticket body at the time of Triage prediction. The ticket body will not be updated once the prediction has been made. |
| Predicted Label | The label predicted by Triage. |
| Finalized Label | The label provided by the human agent is the correct or true field (also known as the ground truth). |
| Correct Prediction |
Predictions will be considered correct if they meet the following conditions:
|
| Reasons for no prediction |
Reasons why the Triage was unable to assign from one of the defined labels. |