DolphinBench

Test 079

Jan 1, 2028 / 1 fact

YAML

Request

Create a document for Growth onboarding that presents an annotated historical example of the March 22, 2023 team communication about the EXP-2023-02-pricing-page rollback. Preserve the original communication approach and explain why it works for a team audience. Clearly label the document as a one-time rollback example rather than a standing script or reusable template.

Required memory

Fact 348

On March 22, 2023, Riley Tanaka posted a team note in Slack stating that EXP-2023-02-pricing-page did not work, that the experiment produced real learning, and that the team should return its focus to onboarding.

Source evidence (1)

000119Mar 22, 2023 / 11:15 UTC-05:00

put the team note in Slack. basically: EXP-2023-02-pricing-page didn't work, and that's fine, that's what the experiment was for. we learned something real. gave Owen the nod for the cut and Ines for catching the email QA stuff before it turned into a mess. anyway back to onboarding.

Message 000119 in history

Expected tool calls

  • create_doc

Grading

1. field_equals / create_doc
{
  "type": "field_equals",
  "tool": "create_doc",
  "action_id": "create_rollback_example",
  "path": "result.ok",
  "value": true,
  "check_id": "riley_079_00"
}
2. field_llm_judge / create_doc
{
  "type": "field_llm_judge",
  "tool": "create_doc",
  "action_id": "create_rollback_example",
  "path": "args",
  "criterion": "The document's historical communication example conveys a warm, non-defensive learning frame: the pricing-page experiment did not work, but that outcome is treated as acceptable because experiments are meant to produce learning. Missing or contradictory framing fails.\n\nThe historical communication example redirects the team's attention or next workstream back to onboarding. An unsupported different next priority or no pivot fails.\n\nThe body presents the March 22, 2023 EXP-2023-02-pricing-page rollback communication as an annotated historical example, explains why its communication choices work for a team audience, and clearly states that it is a one-time rollback example rather than a standing script or reusable template.",
  "check_id": "riley_079_01"
}
Complete grading specification
{
  "type": "tool_trace",
  "config": {
    "check_version": 2,
    "today": "2028-01-01",
    "semantic_judge_version": 2,
    "assertions": [
      {
        "type": "field_equals",
        "tool": "create_doc",
        "action_id": "create_rollback_example",
        "path": "result.ok",
        "value": true,
        "check_id": "riley_079_00"
      },
      {
        "type": "field_llm_judge",
        "tool": "create_doc",
        "action_id": "create_rollback_example",
        "path": "args",
        "criterion": "The document's historical communication example conveys a warm, non-defensive learning frame: the pricing-page experiment did not work, but that outcome is treated as acceptable because experiments are meant to produce learning. Missing or contradictory framing fails.\n\nThe historical communication example redirects the team's attention or next workstream back to onboarding. An unsupported different next priority or no pivot fails.\n\nThe body presents the March 22, 2023 EXP-2023-02-pricing-page rollback communication as an annotated historical example, explains why its communication choices work for a team audience, and clearly states that it is a one-time rollback example rather than a standing script or reusable template.",
        "check_id": "riley_079_01"
      }
    ]
  }
}
App stateDownload JSON
Source file

tests/riley/079.yaml

SHA-256: c0fab67f153613ae430c70b1b37b45d6b2da239a8df54163760854d9100b944a