id stringlengths 6 9 | category stringlengths 3 25 | name stringlengths 12 54 | household unknown | question stringlengths 35 132 | expected dict | scoring unknown | reference_keys listlengths 0 2 | computation stringlengths 39 143 |
|---|---|---|---|---|---|---|---|---|
nw-001 | net_worth | Net worth from the balance sheet | {
"filing_status": "single",
"age": 45,
"total_assets_usd": 850000,
"total_liabilities_usd": 320000
} | What is this household's net worth? | {
"value": 530000,
"unit": "USD"
} | {
"method": "exact"
} | [] | net_worth = total_assets - total_liabilities = 850000 - 320000 = 530000 |
sr-001 | savings_rate | Savings rate | {
"net_income_annual_usd": 96000,
"annual_savings_usd": 19200
} | What is this household's savings rate, as a percent of net income? | {
"value": 20,
"unit": "percent"
} | {
"method": "numeric",
"tolerance_abs": 0.1
} | [] | savings_rate = annual_savings / net_income = 19200 / 96000 = 0.20 = 20% |
ef-001 | emergency_fund | Emergency fund coverage | {
"liquid_assets_usd": 48000,
"monthly_expenses_usd": 8000
} | How many months of expenses does this household's emergency fund cover? | {
"value": 6,
"unit": "months"
} | {
"method": "numeric",
"tolerance_abs": 0.1
} | [] | months = liquid_assets / monthly_expenses = 48000 / 8000 = 6.0 |
dti-001 | debt_to_income | Debt-to-income ratio | {
"monthly_debt_payments_usd": 2400,
"gross_monthly_income_usd": 8000
} | What is this household's debt-to-income (DTI) ratio, as a percent? | {
"value": 30,
"unit": "percent"
} | {
"method": "numeric",
"tolerance_abs": 0.1
} | [] | DTI = monthly_debt_payments / gross_monthly_income = 2400 / 8000 = 0.30 = 30% |
ret-001 | retirement_contribution | Remaining 401(k) elective-deferral room (under 50) | {
"age": 42,
"year": 2026,
"ytd_401k_employee_contribution_usd": 10000
} | How much more can this employee contribute to their 401(k) as an elective deferral in 2026? | {
"value": 14500,
"unit": "USD"
} | {
"method": "exact"
} | [
"retirement.401k.elective_deferral.under_50"
] | 2026 under-50 limit 24500 - already contributed 10000 = 14500 |
ret-002 | retirement_contribution | Maximum 401(k) deferral with age-50 catch-up | {
"age": 58,
"year": 2026
} | What is the maximum total 401(k) elective deferral (including catch-up) for this employee in 2026? | {
"value": 32500,
"unit": "USD"
} | {
"method": "exact"
} | [
"retirement.401k.elective_deferral.under_50",
"retirement.401k.catch_up.age_50_plus"
] | 24500 + 8000 catch-up (age 50+) = 32500 |
ret-003 | retirement_contribution | Maximum 401(k) deferral with SECURE 2.0 super catch-up | {
"age": 61,
"year": 2026
} | What is the maximum total 401(k) elective deferral for this employee in 2026, given the SECURE 2.0 enhanced catch-up for ages 60-63? | {
"value": 35750,
"unit": "USD"
} | {
"method": "exact"
} | [
"retirement.401k.elective_deferral.under_50",
"retirement.401k.catch_up.age_60_63"
] | 24500 + 11250 enhanced catch-up (ages 60-63) = 35750. The 11250 replaces (not adds to) the regular 8000 catch-up. |
rmd-001 | rmd | Required minimum distribution at age 75 | {
"age": 75,
"prior_year_end_ira_balance_usd": 500000
} | What is this person's 2026 required minimum distribution from their IRA, using the Uniform Lifetime Table? | {
"value": 20325.2,
"unit": "USD"
} | {
"method": "numeric",
"tolerance_pct": 0.5
} | [
"retirement.rmd.uniform_lifetime_table"
] | RMD = prior-year-end balance / Uniform Lifetime divisor for age 75 (24.6) = 500000 / 24.6 = 20325.20 |
brk-001 | marginal_tax_bracket | Marginal federal tax bracket | {
"filing_status": "single",
"year": 2026,
"taxable_income_usd": 120000
} | What is this single filer's marginal federal income tax bracket (rate) for 2026? | {
"value": 24,
"unit": "percent"
} | {
"method": "exact"
} | [
"tax.income_brackets.single"
] | 2026 single brackets: 24% applies over 105700 and up to 201775. 120000 falls in the 24% band. |
irmaa-001 | irmaa | Medicare IRMAA tier and total Part B premium | {
"filing_status": "married_filing_jointly",
"magi_2024_usd": 300000
} | For 2026, what is this couple's total monthly Medicare Part B premium per person (standard premium plus IRMAA surcharge)? | {
"value": 405.8,
"unit": "USD/month"
} | {
"method": "numeric",
"tolerance_abs": 0.01
} | [
"medicare.irmaa",
"medicare.part_b.standard_premium_monthly"
] | 2024 MAGI 300000 (MFJ) falls in tier 3 (274000 < MAGI <= 342000): total Part B = 202.90 standard + 202.90 IRMAA surcharge = 405.80 per person. |
est-001 | estate | Estate value exceeding the federal exemption | {
"year": 2026,
"taxable_estate_usd": 20000000
} | How much of this $20,000,000 estate exceeds the 2026 federal estate-tax exemption? | {
"value": 5000000,
"unit": "USD"
} | {
"method": "exact"
} | [
"estate.basic_exclusion"
] | 20000000 - 15000000 (2026 basic exclusion, OBBBA) = 5000000 |
hsa-001 | hsa | Maximum HSA contribution (family, age 50) | {
"age": 50,
"hdhp_coverage": "family",
"year": 2026
} | What is this person's maximum 2026 HSA contribution? | {
"value": 8750,
"unit": "USD"
} | {
"method": "exact"
} | [
"health.hsa.contribution.family",
"health.hsa.catch_up.age_55_plus"
] | Family limit 8750. The age-55 catch-up (1000) does not apply at age 50, so the maximum is 8750. |
st-ca-001 | state_top_rate | California top marginal rate | {
"state": "CA",
"taxable_income_usd": 1500000
} | What is the top marginal California state income tax rate that applies to this resident, including any surtax? | {
"value": 13.3,
"unit": "percent"
} | {
"method": "exact"
} | [
"ca.top_marginal_rate"
] | California 12.3% top bracket + 1% Mental Health Services Tax on income over 1,000,000 = 13.3%. |
st-il-001 | state_retirement_taxation | Illinois tax on a 401(k) distribution | {
"state": "IL",
"annual_401k_distribution_usd": 60000
} | How much Illinois state income tax does this retiree owe on their $60,000 qualified 401(k) distribution? | {
"value": 0,
"unit": "USD"
} | {
"method": "exact"
} | [
"il.retirement_income.state_taxation"
] | Illinois exempts qualified retirement income (401(k) distributions are subtracted on Schedule M), so the Illinois tax on the distribution is 0. |
WealthSchema Planning Benchmark
Is your AI giving 2026 advice — or 2025 advice? Evaluation tasks for AI financial-advice systems, with answer keys built from primary-source-verified U.S. regulatory figure tables — including the enumerated wrong-but-plausible stale values (forbidden figures) with reason codes, so stale answers are countable, not anecdotal.
- v2 (current): 50 tasks — the open split of the AI Eval Sets TY2026
corpus. Rule-grounding and threshold/cliff families, floor/standard
difficulty. Each task carries
answer_key.required_figures(value + unit + establishing source document, e.g. IRS Notice 2025-67) andanswer_key.forbidden_figures(stale/superseded/derived values with reason codes:prior_year_value,superseded,derived_not_published,fabricated_forward_figure, …). Scoring is fully mechanical — no LLM judge. - v1 (superseded, kept for citation continuity): the original 14-task
benchmark with
expected/computationkeys.
Headline metric: the Stale Figure Rate (SFR)
SFR = attempts asserting ≥1 forbidden figure as current ÷ figure-bearing attempts. In the pre-registered pilot of record (40 held-out tasks, 4 systems from 3 labs, k=3), 29% of figure-bearing attempts (86/300) cited a stale or fabricated regulatory figure; per-system SFR ranged 7%–63%, and errors repeated across attempts (pass^3 ≈ pass@1). Full protocol, substitution log, and per-system results: https://www.wealthschema.com/resources/methodology/planning-benchmark-pilot-of-record · live scoreboard: https://www.wealthschema.com/benchmark.
Using it
Blind evaluation: present prompt.system + prompt.user, k=3 attempts;
score mechanically against answer_key (required figures present, forbidden
figures absent as the operative value, typed expectations matched). The live
API serves the same data with a blind mode:
https://www.wealthschema.com/api/benchmark/v2?withhold_answers=true.
Report pass@1, pass^k, and SFR with the denominator, k, and vintage.
Splits and what's deliberately absent
This open split is 50 of a 471-task corpus. The adversarial categorical-flip family, all decision-recall tasks, and a held-out reserve (including all 40 pilot tasks) are never published — that is what keeps the maintainer-run scoreboard re-runnable and third-party comparisons meaningful. The commercial families are sold as one-time eval packs: https://www.wealthschema.com/ai-eval-sets.
License and training use
Free to use for evaluation, benchmarking, and research with attribution
("WealthSchema Planning Benchmark, wealthschema.com/benchmark").
training_use_permitted: false on every record — do not train,
fine-tune, or include these tasks in training corpora; training on
evaluation data destroys its value for everyone it is compared against.
Figures are informational, cited to primary government documents, and are
not tax, legal, or financial advice.
All scenarios are synthetic; no real persons, households, or accounts.
Vintage: each answer key is correct for its stated tax year (vintage,
as_of_date). Figures roll every January; a new vintage of this dataset
ships each year.
Questions: support@capstera.com · https://www.wealthschema.com/for-agents
- Downloads last month
- 31