Datasets:
Tasks:
Text Classification
Modalities:
Text
Formats:
json
Languages:
English
Size:
10K - 100K
License:
id stringlengths 35 47 | state unknown | question stringclasses 607
values | qtype stringclasses 3
values | options listlengths 2 6 | gold int64 0 5 | teacher null | meta dict |
|---|---|---|---|---|---|---|---|
kev:hard-v1/judge/train/00233:correct | "A colleague worked this out.\nAfter a 40% discount, a jacket costs $493.20. What was the price before the discount, to the nearest cent?\nTheir result: Pretty sure it's $822.00\nWorking: $493.20 / (1 - 40%) = $822.00" | Is the colleague's result right? | noul | [
{
"name": "no",
"desc": ""
},
{
"name": "yes",
"desc": ""
}
] | 1 | null | {
"source": "kev_hard_v1",
"family": "kev/judge",
"template": "judge/t1",
"split": "train"
} |
kev:hard-v1/judge/train/00233:value | "A colleague worked this out.\nAfter a 40% discount, a jacket costs $493.20. What was the price before the discount, to the nearest cent?\nTheir result: Pretty sure it's $822.00\nWorking: $493.20 / (1 - 40%) = $822.00" | What should the result be? | choice | [
{
"name": "a",
"desc": "$822.00"
},
{
"name": "b",
"desc": "$690.48"
},
{
"name": "c",
"desc": "$822.01"
},
{
"name": "d",
"desc": "$295.92"
},
{
"name": "e",
"desc": "$533.20"
}
] | 0 | null | {
"source": "kev_hard_v1",
"family": "kev/judge",
"template": "judge/t1",
"split": "train"
} |
kev:hard-v1/ambiguous/train/00660:decision | {
"topic": "Warranty",
"facts": [
"The warranty covers manufacturing defects for 12 months from the purchase date. Accidental damage is never covered.",
"Ines Vance bought a coffee machine on 6 December 2026.",
"Ines made a warranty claim on 3 July 2027.",
"The technician found a faulty heating elem... | What does the rule say should happen? | choice | [
{
"name": "repair_under_warranty",
"desc": "Repair under warranty"
},
{
"name": "insufficient_information",
"desc": "The facts provided do not settle this"
},
{
"name": "not_covered",
"desc": "Not covered by the warranty"
}
] | 0 | null | {
"source": "kev_hard_v1",
"family": "kev/ambiguous",
"template": "ambiguous/t2",
"split": "train"
} |
kev:hard-v1/temporal_numeric/train/03617:date | "TICKET 85894: Recurring schedule\n- The payroll run happens every week on the same weekday.\n- The first run of the year was on Monday 23 February 2026.\n- Question concerns the 8th run of the year, counting the first run as run 1." | What is that date? | choice | [
{
"name": "a",
"desc": "13 April 2026"
},
{
"name": "b",
"desc": "14 April 2026"
},
{
"name": "c",
"desc": "6 April 2026"
},
{
"name": "d",
"desc": "20 April 2026"
}
] | 0 | null | {
"source": "kev_hard_v1",
"family": "kev/temporal_numeric",
"template": "temporal_numeric/t1",
"split": "train"
} |
kev:hard-v1/probability/train/04258:choice | {
"topic": "Project funding decision",
"facts": [
"The planning team must fund exactly one of 3 projects and wants the highest expected net value (expected payoff minus up-front cost).",
"Project Atlas costs $50,000 up front and has a 3-in-10 chance of a gain of $600,000; a 1-in-2 chance of a gain of $380,0... | Based only on these numbers: Which project has the highest expected net value? | choice | [
{
"name": "project_fjord",
"desc": ""
},
{
"name": "project_beacon",
"desc": ""
},
{
"name": "project_atlas",
"desc": ""
}
] | 2 | null | {
"source": "kev_hard_v1",
"family": "kev/probability",
"template": "probability/t2",
"split": "train"
} |
kev:hard-v1/probability/train/04258:value | {
"topic": "Project funding decision",
"facts": [
"The planning team must fund exactly one of 3 projects and wants the highest expected net value (expected payoff minus up-front cost).",
"Project Atlas costs $50,000 up front and has a 3-in-10 chance of a gain of $600,000; a 1-in-2 chance of a gain of $380,0... | Based only on these numbers: What is the expected net value of Project Beacon? | choice | [
{
"name": "a",
"desc": "$290,000"
},
{
"name": "b",
"desc": "$350,000"
},
{
"name": "c",
"desc": "$230,000"
},
{
"name": "d",
"desc": "$396,666"
},
{
"name": "e",
"desc": "$760,000"
}
] | 0 | null | {
"source": "kev_hard_v1",
"family": "kev/probability",
"template": "probability/t2",
"split": "train"
} |
kev:hard-v1/temporal_numeric/train/03445:due | "TICKET 87128: Complaint response deadline\n- Cobalt Ridge must answer every formal complaint within 7 business days of receiving it.\n- Business days are Monday to Friday, excluding the public holidays listed here; the day the complaint is received does not count.\n- Public holidays this year: Heritage Day (Tuesday 11... | What is the deadline? | choice | [
{
"name": "a",
"desc": "26 May 2027"
},
{
"name": "b",
"desc": "21 May 2027"
},
{
"name": "c",
"desc": "27 May 2027"
},
{
"name": "d",
"desc": "28 May 2027"
},
{
"name": "e",
"desc": "25 May 2027"
}
] | 2 | null | {
"source": "kev_hard_v1",
"family": "kev/temporal_numeric",
"template": "temporal_numeric/t1",
"split": "train"
} |
kev:hard-v1/tradeoff/train/04129:choice | "We need to choose a laptop model for the staff hardware refresh.\nHard requirements: its unit price must be at most $1,890; it must offer three-year on-site warranty.\nEach option has been scored from 1 to 10 on security, ease of use and documentation (higher is better). Among the options that meet every requirement, ... | Which model does the decision rule select? | choice | [
{
"name": "keystone_15",
"desc": ""
},
{
"name": "nova_air",
"desc": ""
},
{
"name": "aero_14",
"desc": ""
}
] | 0 | null | {
"source": "kev_hard_v1",
"family": "kev/tradeoff",
"template": "tradeoff/t1",
"split": "train"
} |
kev:hard-v1/tradeoff/train/04129:meets | "We need to choose a laptop model for the staff hardware refresh.\nHard requirements: its unit price must be at most $1,890; it must offer three-year on-site warranty.\nEach option has been scored from 1 to 10 on security, ease of use and documentation (higher is better). Among the options that meet every requirement, ... | Is Aero 14 compliant with all the hard requirements? | noul | [
{
"name": "no",
"desc": ""
},
{
"name": "yes",
"desc": ""
}
] | 0 | null | {
"source": "kev_hard_v1",
"family": "kev/tradeoff",
"template": "tradeoff/t1",
"split": "train"
} |
kev:hard-v1/ambiguous/train/02492:decision | {
"topic": "Return window",
"facts": [
"Returns are accepted if the return request is made within 14 days of delivery.",
"Lotte Moreau ordered a standing desk on March 24, 2025.",
"The desk was delivered on March 26, 2025.",
"Lotte asked to return it on April 18, 2025.",
"The desk is unused and ... | What does the rule say should happen? | choice | [
{
"name": "insufficient_information",
"desc": "The facts provided do not settle this"
},
{
"name": "reject_return",
"desc": "Reject the return as outside the window"
},
{
"name": "accept_return",
"desc": "Accept the return"
}
] | 1 | null | {
"source": "kev_hard_v1",
"family": "kev/ambiguous",
"template": "ambiguous/t2",
"split": "train"
} |
kev:hard-v1/tradeoff/train/00534:choice | "Hi team,\n\nWe need to choose a marketing agency for the spring campaign. After the calls this week: Northlight Studio quoted fee of $21,500, lead time of 5 weeks, portfolio rating of 9/10, with B2B campaign experience, with in-house video production. Paper Kite quoted fee of $13,000, lead time of 9 weeks, portfolio r... | Which option should the team pick? | choice | [
{
"name": "paper_kite",
"desc": ""
},
{
"name": "northlight_studio",
"desc": ""
},
{
"name": "none_qualifies",
"desc": "None of the options meets every requirement"
},
{
"name": "wildfern_agency",
"desc": ""
}
] | 0 | null | {
"source": "kev_hard_v1",
"family": "kev/tradeoff",
"template": "tradeoff/t2",
"split": "train"
} |
kev:hard-v1/tradeoff/train/00534:meets | "Hi team,\n\nWe need to choose a marketing agency for the spring campaign. After the calls this week: Northlight Studio quoted fee of $21,500, lead time of 5 weeks, portfolio rating of 9/10, with B2B campaign experience, with in-house video production. Paper Kite quoted fee of $13,000, lead time of 9 weeks, portfolio r... | Does Wildfern Agency meet every hard requirement? | noul | [
{
"name": "no",
"desc": ""
},
{
"name": "yes",
"desc": ""
}
] | 0 | null | {
"source": "kev_hard_v1",
"family": "kev/tradeoff",
"template": "tradeoff/t2",
"split": "train"
} |
kev:hard-v1/tradeoff/train/01250:choice | "Hi team,\n\nWe need to choose an office lease for the new Leeds team. After the calls this week: Mill Yard quoted monthly rent of $13,400, average commute of 32 minutes, capacity of 45 desks, without on-site parking, without step-free access. Park Row quoted monthly rent of $6,900, average commute of 26 minutes, capac... | Which option should the team pick? | choice | [
{
"name": "station_house",
"desc": ""
},
{
"name": "park_row",
"desc": ""
},
{
"name": "mill_yard",
"desc": ""
},
{
"name": "canal_wharf",
"desc": ""
}
] | 0 | null | {
"source": "kev_hard_v1",
"family": "kev/tradeoff",
"template": "tradeoff/t2",
"split": "train"
} |
kev:hard-v1/tradeoff/train/01250:meets | "Hi team,\n\nWe need to choose an office lease for the new Leeds team. After the calls this week: Mill Yard quoted monthly rent of $13,400, average commute of 32 minutes, capacity of 45 desks, without on-site parking, without step-free access. Park Row quoted monthly rent of $6,900, average commute of 26 minutes, capac... | Is Canal Wharf compliant with all the hard requirements? | noul | [
{
"name": "no",
"desc": ""
},
{
"name": "yes",
"desc": ""
}
] | 0 | null | {
"source": "kev_hard_v1",
"family": "kev/tradeoff",
"template": "tradeoff/t2",
"split": "train"
} |
kev:hard-v1/probability/train/02057:value | "TICKET 91121: Quarterly breakdown\n- Counts of support tickets last quarter by channel and outcome:\n- chat: 399 escalated, 314 not escalated.\n- email: 282 escalated, 348 not escalated." | Using the figures above: Among records that were escalated, what share came from email? | choice | [
{
"name": "a",
"desc": "46.9%"
},
{
"name": "b",
"desc": "21.0%"
},
{
"name": "c",
"desc": "44.8%"
},
{
"name": "d",
"desc": "50.7%"
},
{
"name": "e",
"desc": "41.4%"
}
] | 4 | null | {
"source": "kev_hard_v1",
"family": "kev/probability",
"template": "probability/t1",
"split": "train"
} |
kev:hard-v1/ambiguous/train/03058:decision | "TICKET 78004: Shipping\n- Orders ship free when the order total after all discounts is at least $150.00; otherwise the standard shipping fee applies.\n- Nia Nguyen's basket totals $137.63 before discounts.\n- The coupon Nia applied took $25.00 off.\n- The basket contains three items from the same warehouse.\n- Nia cho... | How should this case be decided under the rule? | choice | [
{
"name": "cannot_determine",
"desc": "Not enough information to decide"
},
{
"name": "standard_shipping",
"desc": "Charge the standard shipping fee"
},
{
"name": "free_shipping",
"desc": "Ship the order free"
}
] | 1 | null | {
"source": "kev_hard_v1",
"family": "kev/ambiguous",
"template": "ambiguous/t1",
"split": "train"
} |
kev:hard-v1/ambiguous/train/02115:decision | "TICKET 17213: Shipping\n- Orders ship free when the order total after all discounts is at least $150.00; otherwise the standard shipping fee applies.\n- Ashok Rahman's basket totals $104.32 before discounts.\n- The basket contains three items from the same warehouse.\n- Ashok chose standard delivery." | How should this case be decided under the rule? | choice | [
{
"name": "cannot_determine",
"desc": "Not enough information to decide"
},
{
"name": "free_shipping",
"desc": "Ship the order free"
},
{
"name": "standard_shipping",
"desc": "Charge the standard shipping fee"
}
] | 2 | null | {
"source": "kev_hard_v1",
"family": "kev/ambiguous",
"template": "ambiguous/t1",
"split": "train"
} |
kev:hard-v1/ambiguous/train/04084:decision | {
"topic": "Overtime",
"facts": [
"Hours beyond 40 in a Monday-to-Sunday week are overtime. No shift may exceed 12 hours.",
"Jonas Rahman worked 8 hours on Monday.",
"Jonas Rahman worked 10 hours on Tuesday.",
"Jonas Rahman worked 6 hours on Wednesday.",
"Jonas Rahman worked 7 hours on Thursday.... | What does the rule say should happen? | choice | [
{
"name": "overtime_due",
"desc": "Overtime is due for this week"
},
{
"name": "insufficient_information",
"desc": "The facts provided do not settle this"
},
{
"name": "no_overtime",
"desc": "No overtime is due for this week"
}
] | 2 | null | {
"source": "kev_hard_v1",
"family": "kev/ambiguous",
"template": "ambiguous/t2",
"split": "train"
} |
kev:hard-v1/multi_hop/train/01994:approver | {
"topic": "Approval routing",
"facts": [
"Purchase requests of this size must be approved by the nearest person above the requester in the reporting line whose title is Vice President or more senior. Seniority, from junior to senior: Analyst, Engineer, Manager, Senior Manager, Director, Vice President, Chief E... | Who is the required approver? | choice | [
{
"name": "wen",
"desc": ""
},
{
"name": "hana",
"desc": ""
},
{
"name": "viktor",
"desc": ""
},
{
"name": "beatriz",
"desc": ""
},
{
"name": "nadia",
"desc": ""
}
] | 2 | null | {
"source": "kev_hard_v1",
"family": "kev/multi_hop",
"template": "multi_hop/t2",
"split": "train"
} |
kev:hard-v1/probability/train/03760:value | "Subject: Alert precision\n\nHi all,\n\nAn inspection camera flags units on the production line. Historically, 40% of units are defective. When one of the units is defective, the system flags it 80% of the time. When one of the units is not defective, the system still flags it 15% of the time.\n\nThanks." | What is the probability that one of the units that the system has flagged is actually defective? | choice | [
{
"name": "a",
"desc": "80.0%"
},
{
"name": "b",
"desc": "40.0%"
},
{
"name": "c",
"desc": "78.0%"
},
{
"name": "d",
"desc": "32.0%"
},
{
"name": "e",
"desc": "85.0%"
}
] | 2 | null | {
"source": "kev_hard_v1",
"family": "kev/probability",
"template": "probability/t0",
"split": "train"
} |
kev:hard-v1/probability/train/03760:bucket | "Subject: Alert precision\n\nHi all,\n\nAn inspection camera flags units on the production line. Historically, 40% of units are defective. When one of the units is defective, the system flags it 80% of the time. When one of the units is not defective, the system still flags it 15% of the time.\n\nThanks." | How likely is it that one of the units that the system has flagged is defective? | score | [
{
"name": "0",
"desc": "Under 10%"
},
{
"name": "1",
"desc": "10% to 30%"
},
{
"name": "2",
"desc": "30% to 50%"
},
{
"name": "3",
"desc": "50% to 70%"
},
{
"name": "4",
"desc": "70% to 90%"
},
{
"name": "5",
"desc": "Over 90%"
}
] | 4 | null | {
"source": "kev_hard_v1",
"family": "kev/probability",
"template": "probability/t0",
"split": "train"
} |
kev:hard-v1/long_policy/train/01461:outcome | {
"policy": "HOME CONTENTS INSURANCE POLICY\nIssued by Lantern General Insurance\nIn this policy, \"the Insurer\" means Lantern General Insurance and \"the Policyholder\" means the person named in the schedule.\n\nArticle 1. Definitions\n(a) Art and antiques means paintings, antiques, rugs and other works of art.\n(b... | What is the correct decision on this claim under the policy? | choice | [
{
"name": "pay_less_fee",
"desc": "Pay the assessed loss less the excess"
},
{
"name": "pay_sublimit",
"desc": "Pay up to the Portable electronics sublimit (Article 3(c))"
},
{
"name": "deny_wear_home",
"desc": "Decline under the wear and tear exclusion (Article 3(e))"
},
{
"... | 1 | null | {
"source": "kev_hard_v1",
"family": "kev/long_policy",
"template": "long_policy/t1",
"split": "train"
} |
kev:hard-v1/long_policy/train/01461:amount | {
"policy": "HOME CONTENTS INSURANCE POLICY\nIssued by Lantern General Insurance\nIn this policy, \"the Insurer\" means Lantern General Insurance and \"the Policyholder\" means the person named in the schedule.\n\nArticle 1. Definitions\n(a) Art and antiques means paintings, antiques, rugs and other works of art.\n(b... | What amount is payable under the policy? | choice | [
{
"name": "a",
"desc": "$500"
},
{
"name": "b",
"desc": "$400"
},
{
"name": "c",
"desc": "$2,900"
},
{
"name": "d",
"desc": "$1,513"
},
{
"name": "e",
"desc": "$1,413"
}
] | 0 | null | {
"source": "kev_hard_v1",
"family": "kev/long_policy",
"template": "long_policy/t1",
"split": "train"
} |
kev:hard-v1/judge/train/03940:correct | "Question: Rule: a discount applies if both the customer used a promo code and the item is on clearance, or if the order total is over $100. Facts: the customer used a promo code; it is not true that the item is on clearance; it is not true that the order total is over $100. Does the discount apply?\nProposed answer: N... | Is the proposed answer correct? | noul | [
{
"name": "no",
"desc": ""
},
{
"name": "yes",
"desc": ""
}
] | 1 | null | {
"source": "kev_hard_v1",
"family": "kev/judge",
"template": "judge/t0",
"split": "train"
} |
kev:hard-v1/probability/train/03185:choice | "TICKET 61796: Comparing odds\n- There is a probability of 0.6 that the shipment clears customs on the first day.\n- There is a probability of 0.25 that the ad campaign beats its target; these two are independent.\n- There is a probability of 0.69 that the audit finds no issues." | Using the figures above: Which is more likely? | choice | [
{
"name": "first_event",
"desc": "At least one of these happens: the shipment clears customs on the first day, or the ad campaign beats its target"
},
{
"name": "equally_likely",
"desc": "They are equally likely"
},
{
"name": "second_event",
"desc": "That the audit finds no issues"
... | 0 | null | {
"source": "kev_hard_v1",
"family": "kev/probability",
"template": "probability/t1",
"split": "train"
} |
kev:hard-v1/long_policy/train/02641:outcome | {
"policy": "TRAVEL INSURANCE POLICY WORDING\nIssued by Meridian Travel Insurance Ltd\nIn this policy, \"the Insurer\" means Meridian Travel Insurance Ltd and \"the Insured Person\" means the person named in the schedule.\n\nArticle 1. Definitions\n(a) Assessed loss means the cost of repair or replacement agreed by t... | What is the correct decision on this claim under the policy? | choice | [
{
"name": "pay_sublimit",
"desc": "Pay up to the Dental treatment sublimit (Article 29(b))"
},
{
"name": "not_covered",
"desc": "Decline: Emergency Medical Expenses is not a cover shown in the schedule"
},
{
"name": "deny_late",
"desc": "Decline: the claim was notified too late (Arti... | 3 | null | {
"source": "kev_hard_v1",
"family": "kev/long_policy",
"template": "long_policy/t1",
"split": "train"
} |
kev:hard-v1/long_policy/train/02641:amount | {
"policy": "TRAVEL INSURANCE POLICY WORDING\nIssued by Meridian Travel Insurance Ltd\nIn this policy, \"the Insurer\" means Meridian Travel Insurance Ltd and \"the Insured Person\" means the person named in the schedule.\n\nArticle 1. Definitions\n(a) Assessed loss means the cost of repair or replacement agreed by t... | What amount is payable under the policy? | choice | [
{
"name": "a",
"desc": "$25,000"
},
{
"name": "b",
"desc": "$0"
},
{
"name": "c",
"desc": "$24,209"
},
{
"name": "d",
"desc": "$23,959"
},
{
"name": "e",
"desc": "$24,750"
}
] | 1 | null | {
"source": "kev_hard_v1",
"family": "kev/long_policy",
"template": "long_policy/t1",
"split": "train"
} |
kev:hard-v1/long_policy/train/02032:outcome | "ELECTRONICS CARE PLAN\nIssued by Tessera Warranty Company\n\nSECTION 1 - DEFINITIONS\n1.1 Accessories means cases, chargers, earbuds and styluses.\n1.2 Accessories means cases, chargers and earbuds.\n1.3 Assessed loss means the cost of repair or replacement agreed by us, before any service fee, limit or sublimit is ap... | Under the policy wording, how should this claim be settled? | choice | [
{
"name": "not_covered",
"desc": "Decline: Accidental Damage is not a cover shown in the schedule"
},
{
"name": "pay_limit",
"desc": "Pay up to the Accidental Damage limit"
},
{
"name": "pay_sublimit",
"desc": "Pay up to the Accessories sublimit (Section 5.2)"
},
{
"name": "d... | 1 | null | {
"source": "kev_hard_v1",
"family": "kev/long_policy",
"template": "long_policy/t0",
"split": "train"
} |
kev:hard-v1/long_policy/train/02032:amount | "ELECTRONICS CARE PLAN\nIssued by Tessera Warranty Company\n\nSECTION 1 - DEFINITIONS\n1.1 Accessories means cases, chargers, earbuds and styluses.\n1.2 Accessories means cases, chargers and earbuds.\n1.3 Assessed loss means the cost of repair or replacement agreed by us, before any service fee, limit or sublimit is ap... | How much should be paid on this claim? | choice | [
{
"name": "a",
"desc": "$1,200"
},
{
"name": "b",
"desc": "$1,889"
},
{
"name": "c",
"desc": "$0"
},
{
"name": "d",
"desc": "$1,790"
},
{
"name": "e",
"desc": "$1,101"
}
] | 0 | null | {
"source": "kev_hard_v1",
"family": "kev/long_policy",
"template": "long_policy/t0",
"split": "train"
} |
kev:hard-v1/probability/train/01727:value | "Engineer: A checkout succeeds only if all 3 independent steps succeed: the payment gateway, the fraud check, the address validator.\nEngineer: The payment gateway succeeds 97% of the time.\nAnalyst: noted.\nEngineer: The fraud check succeeds 85% of the time.\nEngineer: The address validator succeeds 80% of the time." | What is the probability that a checkout fails? | choice | [
{
"name": "a",
"desc": "66.0%"
},
{
"name": "b",
"desc": "80.0%"
},
{
"name": "c",
"desc": "99.9%"
},
{
"name": "d",
"desc": "34.0%"
},
{
"name": "e",
"desc": "0.1%"
}
] | 3 | null | {
"source": "kev_hard_v1",
"family": "kev/probability",
"template": "probability/t3",
"split": "train"
} |
kev:hard-v1/probability/train/01727:bucket | "Engineer: A checkout succeeds only if all 3 independent steps succeed: the payment gateway, the fraud check, the address validator.\nEngineer: The payment gateway succeeds 97% of the time.\nAnalyst: noted.\nEngineer: The fraud check succeeds 85% of the time.\nEngineer: The address validator succeeds 80% of the time." | How likely is it that a checkout fails? | score | [
{
"name": "0",
"desc": "Under 10%"
},
{
"name": "1",
"desc": "10% to 30%"
},
{
"name": "2",
"desc": "30% to 50%"
},
{
"name": "3",
"desc": "50% to 70%"
},
{
"name": "4",
"desc": "70% to 90%"
},
{
"name": "5",
"desc": "Over 90%"
}
] | 2 | null | {
"source": "kev_hard_v1",
"family": "kev/probability",
"template": "probability/t3",
"split": "train"
} |
kev:hard-v1/multi_hop/train/02337:affected | "TICKET 92365: Outage: auth-service\n- Outages propagate along hard dependencies: if a service goes down, every service that depends on it, directly or through other hard dependencies, is affected. Soft dependencies do not propagate outages.\n- session-cache depends on auth-service.\n- inventory-db depends on fraud-sco... | Which listed service is impacted? | choice | [
{
"name": "inventory_db",
"desc": ""
},
{
"name": "billing_api",
"desc": ""
},
{
"name": "user_profile",
"desc": ""
},
{
"name": "fraud_scorer",
"desc": ""
}
] | 2 | null | {
"source": "kev_hard_v1",
"family": "kev/multi_hop",
"template": "multi_hop/t1",
"split": "train"
} |
kev:hard-v1/temporal_numeric/train/04141:refund | "TICKET 7214: Subscription cancellation\n- A customer paid $3,650.00 for an annual plan running from 12 June 2025 up to (but not including) 12 June 2026.\n- They cancelled on 3 November 2025.\n- Refund rule: refund the annual price multiplied by the number of unused days and divided by the number of days in the plan ye... | How much should be refunded? | choice | [
{
"name": "a",
"desc": "$1,450.00"
},
{
"name": "b",
"desc": "$2,210.00"
},
{
"name": "c",
"desc": "$2,129.17"
},
{
"name": "d",
"desc": "$2,200.00"
},
{
"name": "e",
"desc": "$2,193.99"
}
] | 3 | null | {
"source": "kev_hard_v1",
"family": "kev/temporal_numeric",
"template": "temporal_numeric/t1",
"split": "train"
} |
kev:hard-v1/long_policy/train/02528:outcome | "ANNUAL MULTI-TRIP TRAVEL COVER\nIssued by Meridian Travel Insurance Ltd\n\nSECTION 1 - DEFINITIONS\n1.1 Assessed loss means the cost of repair or replacement agreed by us, before any excess, limit or sublimit is applied.\n1.2 Business day means Monday to Friday, excluding public holidays.\n1.3 Dental treatment means t... | Under the policy wording, how should this claim be settled? | choice | [
{
"name": "deny_late",
"desc": "Decline: the claim was notified too late (Section 8.2)"
},
{
"name": "deny_waiting",
"desc": "Decline: the incident happened during the waiting period (Section 5.2)"
},
{
"name": "not_covered",
"desc": "Decline: Travel Delay is not a cover shown in the... | 0 | null | {
"source": "kev_hard_v1",
"family": "kev/long_policy",
"template": "long_policy/t0",
"split": "train"
} |
kev:hard-v1/long_policy/train/02528:amount | "ANNUAL MULTI-TRIP TRAVEL COVER\nIssued by Meridian Travel Insurance Ltd\n\nSECTION 1 - DEFINITIONS\n1.1 Assessed loss means the cost of repair or replacement agreed by us, before any excess, limit or sublimit is applied.\n1.2 Business day means Monday to Friday, excluding public holidays.\n1.3 Dental treatment means t... | How much should be paid on this claim? | choice | [
{
"name": "a",
"desc": "$249"
},
{
"name": "b",
"desc": "$274"
},
{
"name": "c",
"desc": "$0"
},
{
"name": "d",
"desc": "$275"
},
{
"name": "e",
"desc": "$300"
}
] | 2 | null | {
"source": "kev_hard_v1",
"family": "kev/long_policy",
"template": "long_policy/t0",
"split": "train"
} |
kev:hard-v1/probability/train/04164:choice | "Subject: Project funding decision\n\nHi all,\n\nThe planning team must fund exactly one of 4 projects and wants the highest expected net value (expected payoff minus up-front cost). Project Cedar costs $70,000 up front and has a 80% chance of a gain of $610,000; a 15% chance of a gain of $500,000; a 5% chance of a los... | Which project has the highest expected net value? | choice | [
{
"name": "project_ember",
"desc": ""
},
{
"name": "project_cedar",
"desc": ""
},
{
"name": "project_beacon",
"desc": ""
},
{
"name": "project_atlas",
"desc": ""
}
] | 1 | null | {
"source": "kev_hard_v1",
"family": "kev/probability",
"template": "probability/t0",
"split": "train"
} |
kev:hard-v1/probability/train/04164:value | "Subject: Project funding decision\n\nHi all,\n\nThe planning team must fund exactly one of 4 projects and wants the highest expected net value (expected payoff minus up-front cost). Project Cedar costs $70,000 up front and has a 80% chance of a gain of $610,000; a 15% chance of a gain of $500,000; a 5% chance of a los... | What is the expected net value of Project Ember? | choice | [
{
"name": "a",
"desc": "$330,000"
},
{
"name": "b",
"desc": "-$187,000"
},
{
"name": "c",
"desc": "-$47,000"
},
{
"name": "d",
"desc": "$93,000"
},
{
"name": "e",
"desc": "$40,000"
}
] | 2 | null | {
"source": "kev_hard_v1",
"family": "kev/probability",
"template": "probability/t0",
"split": "train"
} |
kev:hard-v1/multi_hop/train/00748:paged | "Subject: Alert from card tokenizer\n\nHi all,\n\nPaging rule: an alert pages the owning team's primary on-call; if the primary is on leave that day, it pages the secondary; if the secondary is also on leave, it pages the team's engineering manager. This week's Mobile primary on-call is Tess and the secondary is Wen. A... | Who gets paged for this alert? | choice | [
{
"name": "wen",
"desc": ""
},
{
"name": "tess",
"desc": ""
},
{
"name": "samir",
"desc": ""
},
{
"name": "sofia",
"desc": ""
}
] | 1 | null | {
"source": "kev_hard_v1",
"family": "kev/multi_hop",
"template": "multi_hop/t0",
"split": "train"
} |
kev:hard-v1/ambiguous/train/00649:decision | "Subject: Price match\n\nHi all,\n\nWe match a competitor's lower price for an identical item if the competitor has it in stock at the time of the request, but we never match a price more than 20% below ours. If the competitor's price is not lower, no adjustment is needed. Amara Haddad asked us to match Oakhaven Bank's... | What is the correct decision? | choice | [
{
"name": "no_adjustment",
"desc": "No adjustment needed"
},
{
"name": "insufficient_information",
"desc": "Cannot be determined from the information given"
},
{
"name": "match_price",
"desc": "Match the competitor's price"
},
{
"name": "decline_match",
"desc": "Decline t... | 1 | null | {
"source": "kev_hard_v1",
"family": "kev/ambiguous",
"template": "ambiguous/t0",
"split": "train"
} |
kev:hard-v1/probability/train/02295:value | "Engineer: An order is confirmed if at least one of 4 redundant suppliers can fill it; each supplier's availability is independent of the others.\nEngineer: Supplier A is able to fill an order 95% of the time.\nAnalyst: noted.\nEngineer: Supplier B is able to fill an order 85% of the time.\nEngineer: Supplier C is able... | What is the probability that at least one supplier can fill the order? | choice | [
{
"name": "a",
"desc": "74.4%"
},
{
"name": "b",
"desc": "100.0%"
},
{
"name": "c",
"desc": "0.0%"
},
{
"name": "d",
"desc": "85.0%"
},
{
"name": "e",
"desc": "25.6%"
}
] | 1 | null | {
"source": "kev_hard_v1",
"family": "kev/probability",
"template": "probability/t3",
"split": "train"
} |
kev:hard-v1/probability/train/02295:bucket | "Engineer: An order is confirmed if at least one of 4 redundant suppliers can fill it; each supplier's availability is independent of the others.\nEngineer: Supplier A is able to fill an order 95% of the time.\nAnalyst: noted.\nEngineer: Supplier B is able to fill an order 85% of the time.\nEngineer: Supplier C is able... | How likely is it that at least one supplier can fill the order? | score | [
{
"name": "0",
"desc": "Under 10%"
},
{
"name": "1",
"desc": "10% to 30%"
},
{
"name": "2",
"desc": "30% to 50%"
},
{
"name": "3",
"desc": "50% to 70%"
},
{
"name": "4",
"desc": "70% to 90%"
},
{
"name": "5",
"desc": "Over 90%"
}
] | 5 | null | {
"source": "kev_hard_v1",
"family": "kev/probability",
"template": "probability/t3",
"split": "train"
} |
kev:hard-v1/multi_hop/train/01569:paged | "TICKET 37747: Alert from card tokenizer\n- Paging rule: an alert pages the owning team's primary on-call; if the primary is on leave that day, it pages the secondary; if the secondary is also on leave, it pages the team's engineering manager.\n- Alerts from the cart service are owned by the Search team.\n- Alerts from... | Which person should be paged? | choice | [
{
"name": "kofi",
"desc": ""
},
{
"name": "tomas",
"desc": ""
},
{
"name": "nadia",
"desc": ""
},
{
"name": "esme",
"desc": ""
}
] | 2 | null | {
"source": "kev_hard_v1",
"family": "kev/multi_hop",
"template": "multi_hop/t1",
"split": "train"
} |
kev:hard-v1/long_policy/train/00313:outcome | {
"policy": "GROUP HEALTH CASH PLAN\nIssued by the Staff Welfare Fund\nIn this policy, \"the Plan\" means the Staff Welfare Fund and \"the Member\" means the person named in the schedule.\n\nArticle 1. Definitions\n(a) Hearing accessories means batteries, chargers and cleaning kits.\n(b) Hearing aid replacement perio... | What is the correct decision on this claim under the policy? | choice | [
{
"name": "pay_sublimit",
"desc": "Pay up to the Hearing accessories sublimit (Article 3(c))"
},
{
"name": "pay_limit",
"desc": "Pay up to the Hearing Care limit"
},
{
"name": "pay_less_fee",
"desc": "Pay the assessed loss less the deductible"
},
{
"name": "not_covered",
... | 3 | null | {
"source": "kev_hard_v1",
"family": "kev/long_policy",
"template": "long_policy/t1",
"split": "train"
} |
kev:hard-v1/long_policy/train/00313:amount | {
"policy": "GROUP HEALTH CASH PLAN\nIssued by the Staff Welfare Fund\nIn this policy, \"the Plan\" means the Staff Welfare Fund and \"the Member\" means the person named in the schedule.\n\nArticle 1. Definitions\n(a) Hearing accessories means batteries, chargers and cleaning kits.\n(b) Hearing aid replacement perio... | What amount is payable under the policy? | choice | [
{
"name": "a",
"desc": "$1,574.40"
},
{
"name": "b",
"desc": "$2,000"
},
{
"name": "c",
"desc": "$0"
},
{
"name": "d",
"desc": "$1,900"
},
{
"name": "e",
"desc": "$1,474.40"
}
] | 2 | null | {
"source": "kev_hard_v1",
"family": "kev/long_policy",
"template": "long_policy/t1",
"split": "train"
} |
kev:hard-v1/temporal_numeric/train/02299:days | "Finance: Payment reached the account on June 26, 2025.\nFinance: Invoice 64161 was issued on May 18, 2025.\nAnalyst: ok, go on.\nFinance: A late fee applies when payment arrives more than 30 days after the invoice date." | What is the gap in days between invoice and payment? | choice | [
{
"name": "a",
"desc": "40 days"
},
{
"name": "b",
"desc": "41 days"
},
{
"name": "c",
"desc": "39 days"
},
{
"name": "d",
"desc": "38 days"
},
{
"name": "e",
"desc": "37 days"
}
] | 2 | null | {
"source": "kev_hard_v1",
"family": "kev/temporal_numeric",
"template": "temporal_numeric/t3",
"split": "train"
} |
kev:hard-v1/temporal_numeric/train/02299:late | "Finance: Payment reached the account on June 26, 2025.\nFinance: Invoice 64161 was issued on May 18, 2025.\nAnalyst: ok, go on.\nFinance: A late fee applies when payment arrives more than 30 days after the invoice date." | Does the late fee apply? | noul | [
{
"name": "no",
"desc": ""
},
{
"name": "yes",
"desc": ""
}
] | 1 | null | {
"source": "kev_hard_v1",
"family": "kev/temporal_numeric",
"template": "temporal_numeric/t3",
"split": "train"
} |
kev:hard-v1/temporal_numeric/train/02681:date | "TICKET 29915: Recurring schedule\n- Invoices go out on the last business day (Monday to Friday) of each month.\n- October 31, 2027 is a Sunday.\n- Question concerns the invoice run for October 2027." | What is that date? | choice | [
{
"name": "a",
"desc": "1 November 2027"
},
{
"name": "b",
"desc": "31 October 2027"
},
{
"name": "c",
"desc": "29 October 2027"
},
{
"name": "d",
"desc": "28 October 2027"
},
{
"name": "e",
"desc": "30 October 2027"
}
] | 2 | null | {
"source": "kev_hard_v1",
"family": "kev/temporal_numeric",
"template": "temporal_numeric/t1",
"split": "train"
} |
kev:hard-v1/judge/train/03421:correct | "A colleague worked this out.\nConvert 28 degrees Celsius to degrees Fahrenheit (F = C x 9/5 + 32). Give the answer to one decimal place.\nTheir result: I think 108.0 degrees Fahrenheit\nWorking: 28 x 1.8 + 32 = 108.0" | Is the colleague's result right? | noul | [
{
"name": "no",
"desc": ""
},
{
"name": "yes",
"desc": ""
}
] | 0 | null | {
"source": "kev_hard_v1",
"family": "kev/judge",
"template": "judge/t1",
"split": "train"
} |
kev:hard-v1/judge/train/03421:value | "A colleague worked this out.\nConvert 28 degrees Celsius to degrees Fahrenheit (F = C x 9/5 + 32). Give the answer to one decimal place.\nTheir result: I think 108.0 degrees Fahrenheit\nWorking: 28 x 1.8 + 32 = 108.0" | What should the result be? | choice | [
{
"name": "a",
"desc": "824.0 degrees Fahrenheit"
},
{
"name": "b",
"desc": "47.6 degrees Fahrenheit"
},
{
"name": "c",
"desc": "108.0 degrees Fahrenheit"
},
{
"name": "d",
"desc": "50.4 degrees Fahrenheit"
},
{
"name": "e",
"desc": "82.4 degrees Fahrenheit"
... | 4 | null | {
"source": "kev_hard_v1",
"family": "kev/judge",
"template": "judge/t1",
"split": "train"
} |
kev:hard-v1/judge/train/01936:correct | "Question: Monthly active users went from 2,477 to 2,421. What is the percentage change, to one decimal place (negative for a decrease)?\nProposed answer: Final answer: -56.0%" | Is the proposed answer correct? | noul | [
{
"name": "no",
"desc": ""
},
{
"name": "yes",
"desc": ""
}
] | 0 | null | {
"source": "kev_hard_v1",
"family": "kev/judge",
"template": "judge/t0",
"split": "train"
} |
kev:hard-v1/judge/train/01936:value | "Question: Monthly active users went from 2,477 to 2,421. What is the percentage change, to one decimal place (negative for a decrease)?\nProposed answer: Final answer: -56.0%" | What is the correct answer? | choice | [
{
"name": "a",
"desc": "-56.0%"
},
{
"name": "b",
"desc": "97.7%"
},
{
"name": "c",
"desc": "-2.3%"
},
{
"name": "d",
"desc": "2.3%"
},
{
"name": "e",
"desc": "-55.9%"
}
] | 2 | null | {
"source": "kev_hard_v1",
"family": "kev/judge",
"template": "judge/t0",
"split": "train"
} |
End of preview. Expand in Data Studio
Plumb training data
Training data and split definitions for the Plumb typed-decision models. Every row is in the Plumb Row schema: state, question, qtype (choice | noul | score), options (name, desc), gold (option index), meta.
| Config | Rows | Contents | Redistributed |
|---|---|---|---|
kev_hard |
46,533 train · 1,546 development · 1,550 test | Solver-labelled hard-skill decisions in seven families (long_policy, multi_hop, judge, tradeoff, probability, temporal_numeric, ambiguous), generated with Kev's build_hard_v1 (Apache-2.0). Development and test use surface templates held out of training. |
rows |
decisionbench_ood_v0 |
15,242 train · 1,692 dev_in · 6,966 test_ood | The leave-families-out split of DecisionBench used for Plumb v1/v2: six held-out families (routing_triage, guardrails_moderation, document_workflows, risk_scoring, triage, content_moderation). | row ids + family folds; rows rebuild from DecisionBench (license: other) |
ext_v2 |
437,767 rows over 318 tasks / 195 datasets | The tasksource classification breadth set used for Plumb v2, after the provenance denylist, with 400 rows/task and 1,500 rows/dataset caps in the mix. | task manifest + the 129,677 v2 mix ids; rows rebuild from the upstream datasets (mixed licenses) |
Rebuild
pip install git+https://github.com/therealnaveenkamal/plumber
python -m plumber.data.decisionbench --parquet <DecisionBench eval parquet> --out data/decisionbench
python -m plumber.data.split --rows data/decisionbench/rows.jsonl --out data/ood_v0 \
--holdout routing_triage guardrails_moderation document_workflows risk_scoring triage content_moderation
python -m plumber.data.tasksource --out data/ext_v2 --cap 1500
python -m plumber.data.screen --candidates data/ext_v2/rows.jsonl --heldout data/ood_v0/test_ood.jsonl --out data/ext_v2/rows_clean.jsonl
python -m plumber.data.mix --ext data/ext_v2/rows_clean.jsonl --cap_per_task 400 --cap_per_dataset 1500 --db data/ood_v0/train.jsonl --db_repeat 2 --out data/v2/train.jsonl
The split and mix are deterministic (seed 20260927); decisionbench_ood_v0/*_ids.json and ext_v2/v2_mix_ids.json are the exact row ids used.
Contamination
kev_hard and ext_v2 were screened against the six held-out DecisionBench families and against all 231 JevBench public items (exact normalized text and 8-gram Jaccard ≥ 0.2): 1 row removed from ext_v2, 0 from kev_hard. The denylist of DecisionBench upstream datasets and their siblings applied to ext_v2 is in plumber/data/tasksource.py.
License
kev_hard rows and all split/manifest files: Apache-2.0. Rebuilt rows carry their upstream licenses.
- Downloads last month
- 67