desktop

Navigate

Overview DesktopBench Leaderboard RL Envs Evaluation
Main site Evals Blog Dashboard
core candidate

SupportOps-200

Support workflows across ticket classification, replies, macros, customer lookup, order lookup, escalations, follow-ups, and closure.

Core release candidate: SupportOps-50

Task families

20 families x 10 variants

TicketClassifyTicketAssignReplyDraftRefundEscalationStatusUpdatePriorityChangeSLAReviewMacroApplyKnowledgeBaseSearchCustomerLookupOrderLookupBugReportCreateDuplicateTicketMergeTagAddSentimentFlagAttachmentReviewFollowUpScheduleCloseTicketReopenTicketCSATReview
Target UIs

Mock apps for resettable rollouts

Zendesk-like ticket desk

Intercom-like inbox

Freshdesk-like queue

Gmail-like support mailbox

Examples

Representative task contracts

Example 1

SupportOps.TicketClassify.001

Instruction
Classify ticket #5831 as Billing and set priority to High.
Grader
ticket.category == Billing and ticket.priority == High
Example 2

SupportOps.OrderLookup.001

Instruction
Customer asks about order #1071. Check the order system and reply with the current tracking number.
Grader
reply includes tracking number from order #1071 and is sent to the correct ticket
Example 3

SupportOps.MacroApply.001

Instruction
Apply the Delayed Shipping Apology macro to ticket #5882 and personalize it with the customer first name.
Grader
macro_id == delayed_shipping_apology and reply contains correct first name
Verifier

Suite-level verifier shape

Ticket DB state, message body checks, tag/assignee/priority checks, and public-vs-internal reply guards.

On this page

  • Summary
  • Task families
  • Examples
  • Verifier

Help improve UseDesktop

Optional analytics help us learn which pages are useful. They stay off unless you accept.

Privacy