DolphinBench

Test 190

Jan 1, 2028 / 2 facts

YAML

Request

Create a historical decision card for EXP-2023-02-onboarding-trigger at the point when Riley called the experiment successful and began planning full rollout.

Required memory

Fact 352

Riley Tanaka decided to call EXP-2023-02-onboarding-trigger as successful after it cleared the second cohort.

Source evidence (2)

000147Apr 3, 2023 / 14:00 UTC-05:00

ok so EXP-2023-02-onboarding-trigger cleared on the second cohort too. pooled week-1 retention is +3.1 pts treatment vs control, p=0.04, n=420 across the two cohorts. I called it. still doing the boring checks with Owen's pull -- SRM, signup-source mix, any weird country/device skew -- but unless something dumb pops up we're drafting the ship plan with Ines now and putting full rollout at 100% on Apr 10. nice clean one after the pricing-page mess lol

Message 000147 in history
004079Mar 5, 2027 / 08:05 UTC-06:00

Send a text-only email to the electricity supplier with subject `Unexpected $9.95 plan service fee on Cherrywood electricity statement` and this exact body: `Hello, my latest statement includes a separate $9.95 plan service fee. My active fixed plan runs through August 12, 2027 and has no monthly plan fee. The statement otherwise shows 721 kWh at the contracted $0.138 per kWh rate. Please identify the contractual basis for this fee or correct the statement, and confirm the result in writing. Thank you, Riley Tanaka.`

Message 004079 in history

Fact 353

Riley Tanaka and Ines are drafting a ship plan for a 100% rollout of EXP-2023-02-onboarding-trigger on April 10, 2023, unless checks of Owen's pull reveal sample-ratio mismatch, signup-source mix issues, or unusual country/device skew.

Source evidence (1)

000147Apr 3, 2023 / 14:00 UTC-05:00

ok so EXP-2023-02-onboarding-trigger cleared on the second cohort too. pooled week-1 retention is +3.1 pts treatment vs control, p=0.04, n=420 across the two cohorts. I called it. still doing the boring checks with Owen's pull -- SRM, signup-source mix, any weird country/device skew -- but unless something dumb pops up we're drafting the ship plan with Ines now and putting full rollout at 100% on Apr 10. nice clean one after the pricing-page mess lol

Message 000147 in history

Expected tool calls

  • create_doc

Grading

1. field_equals / create_doc
{
  "type": "field_equals",
  "tool": "create_doc",
  "action_id": "create_decision_card",
  "path": "result.ok",
  "value": true,
  "check_id": "riley_190_00"
}
2. field_llm_judge / create_doc
{
  "type": "field_llm_judge",
  "tool": "create_doc",
  "action_id": "create_decision_card",
  "path": "args",
  "criterion": "The card says Riley called EXP-2023-02-onboarding-trigger successful after it cleared the second cohort. It says Riley and Ines were drafting a 100% rollout plan for April 10, 2023, subject to Owen's pull showing no sample-ratio mismatch, signup-source mix issue, or unusual country or device skew. It must distinguish the planned rollout from a completed rollout.",
  "check_id": "riley_190_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_decision_card",
        "path": "result.ok",
        "value": true,
        "check_id": "riley_190_00"
      },
      {
        "type": "field_llm_judge",
        "tool": "create_doc",
        "action_id": "create_decision_card",
        "path": "args",
        "criterion": "The card says Riley called EXP-2023-02-onboarding-trigger successful after it cleared the second cohort. It says Riley and Ines were drafting a 100% rollout plan for April 10, 2023, subject to Owen's pull showing no sample-ratio mismatch, signup-source mix issue, or unusual country or device skew. It must distinguish the planned rollout from a completed rollout.",
        "check_id": "riley_190_01"
      }
    ]
  }
}
App stateDownload JSON
Source file

tests/riley/190.yaml

SHA-256: ce5b11f5943d08603bd2efdb4622c33203960366e08b961076b48b1a142a3440