aarav1656/pf-debt-payoff

CreditCardAPRPath

CreditCardAPRPath is an original, provenance-tracked instruction-tuning dataset for consumer-finance educators and model builders who need payoff advice backed by recomputable arithmetic. Each prompt describes one to five debts and a fixed monthly budget; each label compares debt avalanche with debt snowball month by month and returns exact interest and payoff results. There is no upstream source dataset: the scenarios and cent-exact labels were created together by this project, filling the gap between generic financial-advice text and auditable strategy comparisons with complete numerical paths.

What this data produced

Model: TODO(pf-model) | Adaptive Data quality: 9.0 -> 9.0 (A)

The target failure mode is plausible-sounding payoff advice with wrong or unverifiable numbers. These labels make the arithmetic, assumptions, strategy choice, and disclaimer part of the required answer.

Schema

Each line in train-tagged.jsonl is a JSON object. completion is stored as a JSON-encoded string.

{
  "prompt": "string: task instruction followed by debts, APRs, minimums, and monthly budget",
  "completion": "string containing the payoff-comparison JSON described below",
  "shard": "pf_debt_payoff",
  "license": "cc0-1.0",
  "source": "Original (this work)"
}

After decoding completion, its exact shape is:

{
  "inputs": {
    "debts": [
      {
        "name": "string",
        "balance": 0.0,
        "apr_percent": 0.0,
        "minimum_payment": 0.0
      }
    ],
    "monthly_budget": 0.0
  },
  "month_by_month": {
    "avalanche": {
      "coverage": "string",
      "months": [
        {
          "month": 0,
          "target": "string",
          "opening_balances": {"debt name": 0.0},
          "interest_charged": {"debt name": 0.0},
          "payments": {"debt name": 0.0},
          "ending_balances": {"debt name": 0.0}
        }
      ]
    },
    "snowball": {
      "coverage": "string",
      "months": [
        {
          "month": 0,
          "target": "string",
          "opening_balances": {"debt name": 0.0},
          "interest_charged": {"debt name": 0.0},
          "payments": {"debt name": 0.0},
          "ending_balances": {"debt name": 0.0}
        }
      ]
    }
  },
  "recommended_strategy": "avalanche | snowball | tie",
  "total_interest_avalanche": 0.0,
  "total_interest_snowball": 0.0,
  "months_to_debt_free": {"avalanche": 0, "snowball": 0},
  "assumptions": ["string"],
  "disclaimer": "string",
  "recommendation": "string"
}

The numeric zeros above indicate number-valued fields; they are schema markers, not a fabricated training row.

Real training example

This is the first row of data/pf-debt/train-tagged.jsonl. The completion is decoded only for readability; its values are copied from the file.

Show example

{
  "prompt": "You are a personal-finance arithmetic assistant. Compare debt avalanche (highest APR first) with debt snowball (smallest current balance first). Return strict JSON with inputs, month_by_month, recommended_strategy, total_interest_avalanche, total_interest_snowball, months_to_debt_free, assumptions, disclaimer, and recommendation. Show dollar figures to the cent and do not name a financial product or provider.\n\nCONSUMER SITUATION:\nA consumer wants a precise plan for paying off several balances.\nDebts:\n- Debt A: balance $26962.49, APR 1.38%, fixed minimum $1180.96\nFixed total available each month: $1367.00.\nWhich debt should receive extra payments first, and what are the total interest and payoff time under avalanche versus snowball? Show recomputable arithmetic.",
  "completion": {
    "inputs": {
      "debts": [
        {"name": "Debt A", "balance": 26962.49, "apr_percent": 1.38, "minimum_payment": 1180.96}
      ],
      "monthly_budget": 1367.0
    },
    "month_by_month": {
      "avalanche": {
        "coverage": "all months",
        "months": [
          {"month": 1, "target": "Debt A", "opening_balances": {"Debt A": 26962.49}, "interest_charged": {"Debt A": 31.01}, "payments": {"Debt A": 1367.0}, "ending_balances": {"Debt A": 25626.5}},
          {"month": 2, "target": "Debt A", "opening_balances": {"Debt A": 25626.5}, "interest_charged": {"Debt A": 29.47}, "payments": {"Debt A": 1367.0}, "ending_balances": {"Debt A": 24288.97}},
          {"month": 3, "target": "Debt A", "opening_balances": {"Debt A": 24288.97}, "interest_charged": {"Debt A": 27.93}, "payments": {"Debt A": 1367.0}, "ending_balances": {"Debt A": 22949.9}},
          {"month": 4, "target": "Debt A", "opening_balances": {"Debt A": 22949.9}, "interest_charged": {"Debt A": 26.39}, "payments": {"Debt A": 1367.0}, "ending_balances": {"Debt A": 21609.29}},
          {"month": 5, "target": "Debt A", "opening_balances": {"Debt A": 21609.29}, "interest_charged": {"Debt A": 24.85}, "payments": {"Debt A": 1367.0}, "ending_balances": {"Debt A": 20267.14}},
          {"month": 6, "target": "Debt A", "opening_balances": {"Debt A": 20267.14}, "interest_charged": {"Debt A": 23.31}, "payments": {"Debt A": 1367.0}, "ending_balances": {"Debt A": 18923.45}},
          {"month": 7, "target": "Debt A", "opening_balances": {"Debt A": 18923.45}, "interest_charged": {"Debt A": 21.76}, "payments": {"Debt A": 1367.0}, "ending_balances": {"Debt A": 17578.21}},
          {"month": 8, "target": "Debt A", "opening_balances": {"Debt A": 17578.21}, "interest_charged": {"Debt A": 20.21}, "payments": {"Debt A": 1367.0}, "ending_balances": {"Debt A": 16231.42}},
          {"month": 9, "target": "Debt A", "opening_balances": {"Debt A": 16231.42}, "interest_charged": {"Debt A": 18.67}, "payments": {"Debt A": 1367.0}, "ending_balances": {"Debt A": 14883.09}},
          {"month": 10, "target": "Debt A", "opening_balances": {"Debt A": 14883.09}, "interest_charged": {"Debt A": 17.12}, "payments": {"Debt A": 1367.0}, "ending_balances": {"Debt A": 13533.21}},
          {"month": 11, "target": "Debt A", "opening_balances": {"Debt A": 13533.21}, "interest_charged": {"Debt A": 15.56}, "payments": {"Debt A": 1367.0}, "ending_balances": {"Debt A": 12181.77}},
          {"month": 12, "target": "Debt A", "opening_balances": {"Debt A": 12181.77}, "interest_charged": {"Debt A": 14.01}, "payments": {"Debt A": 1367.0}, "ending_balances": {"Debt A": 10828.78}},
          {"month": 13, "target": "Debt A", "opening_balances": {"Debt A": 10828.78}, "interest_charged": {"Debt A": 12.45}, "payments": {"Debt A": 1367.0}, "ending_balances": {"Debt A": 9474.23}},
          {"month": 14, "target": "Debt A", "opening_balances": {"Debt A": 9474.23}, "interest_charged": {"Debt A": 10.9}, "payments": {"Debt A": 1367.0}, "ending_balances": {"Debt A": 8118.13}},
          {"month": 15, "target": "Debt A", "opening_balances": {"Debt A": 8118.13}, "interest_charged": {"Debt A": 9.34}, "payments": {"Debt A": 1367.0}, "ending_balances": {"Debt A": 6760.47}},
          {"month": 16, "target": "Debt A", "opening_balances": {"Debt A": 6760.47}, "interest_charged": {"Debt A": 7.77}, "payments": {"Debt A": 1367.0}, "ending_balances": {"Debt A": 5401.24}},
          {"month": 17, "target": "Debt A", "opening_balances": {"Debt A": 5401.24}, "interest_charged": {"Debt A": 6.21}, "payments": {"Debt A": 1367.0}, "ending_balances": {"Debt A": 4040.45}},
          {"month": 18, "target": "Debt A", "opening_balances": {"Debt A": 4040.45}, "interest_charged": {"Debt A": 4.65}, "payments": {"Debt A": 1367.0}, "ending_balances": {"Debt A": 2678.1}},
          {"month": 19, "target": "Debt A", "opening_balances": {"Debt A": 2678.1}, "interest_charged": {"Debt A": 3.08}, "payments": {"Debt A": 1367.0}, "ending_balances": {"Debt A": 1314.18}},
          {"month": 20, "target": "Debt A", "opening_balances": {"Debt A": 1314.18}, "interest_charged": {"Debt A": 1.51}, "payments": {"Debt A": 1315.69}, "ending_balances": {"Debt A": 0.0}}
        ]
      },
      "snowball": {
        "coverage": "all months",
        "months": [
          {"month": 1, "target": "Debt A", "opening_balances": {"Debt A": 26962.49}, "interest_charged": {"Debt A": 31.01}, "payments": {"Debt A": 1367.0}, "ending_balances": {"Debt A": 25626.5}},
          {"month": 2, "target": "Debt A", "opening_balances": {"Debt A": 25626.5}, "interest_charged": {"Debt A": 29.47}, "payments": {"Debt A": 1367.0}, "ending_balances": {"Debt A": 24288.97}},
          {"month": 3, "target": "Debt A", "opening_balances": {"Debt A": 24288.97}, "interest_charged": {"Debt A": 27.93}, "payments": {"Debt A": 1367.0}, "ending_balances": {"Debt A": 22949.9}},
          {"month": 4, "target": "Debt A", "opening_balances": {"Debt A": 22949.9}, "interest_charged": {"Debt A": 26.39}, "payments": {"Debt A": 1367.0}, "ending_balances": {"Debt A": 21609.29}},
          {"month": 5, "target": "Debt A", "opening_balances": {"Debt A": 21609.29}, "interest_charged": {"Debt A": 24.85}, "payments": {"Debt A": 1367.0}, "ending_balances": {"Debt A": 20267.14}},
          {"month": 6, "target": "Debt A", "opening_balances": {"Debt A": 20267.14}, "interest_charged": {"Debt A": 23.31}, "payments": {"Debt A": 1367.0}, "ending_balances": {"Debt A": 18923.45}},
          {"month": 7, "target": "Debt A", "opening_balances": {"Debt A": 18923.45}, "interest_charged": {"Debt A": 21.76}, "payments": {"Debt A": 1367.0}, "ending_balances": {"Debt A": 17578.21}},
          {"month": 8, "target": "Debt A", "opening_balances": {"Debt A": 17578.21}, "interest_charged": {"Debt A": 20.21}, "payments": {"Debt A": 1367.0}, "ending_balances": {"Debt A": 16231.42}},
          {"month": 9, "target": "Debt A", "opening_balances": {"Debt A": 16231.42}, "interest_charged": {"Debt A": 18.67}, "payments": {"Debt A": 1367.0}, "ending_balances": {"Debt A": 14883.09}},
          {"month": 10, "target": "Debt A", "opening_balances": {"Debt A": 14883.09}, "interest_charged": {"Debt A": 17.12}, "payments": {"Debt A": 1367.0}, "ending_balances": {"Debt A": 13533.21}},
          {"month": 11, "target": "Debt A", "opening_balances": {"Debt A": 13533.21}, "interest_charged": {"Debt A": 15.56}, "payments": {"Debt A": 1367.0}, "ending_balances": {"Debt A": 12181.77}},
          {"month": 12, "target": "Debt A", "opening_balances": {"Debt A": 12181.77}, "interest_charged": {"Debt A": 14.01}, "payments": {"Debt A": 1367.0}, "ending_balances": {"Debt A": 10828.78}},
          {"month": 13, "target": "Debt A", "opening_balances": {"Debt A": 10828.78}, "interest_charged": {"Debt A": 12.45}, "payments": {"Debt A": 1367.0}, "ending_balances": {"Debt A": 9474.23}},
          {"month": 14, "target": "Debt A", "opening_balances": {"Debt A": 9474.23}, "interest_charged": {"Debt A": 10.9}, "payments": {"Debt A": 1367.0}, "ending_balances": {"Debt A": 8118.13}},
          {"month": 15, "target": "Debt A", "opening_balances": {"Debt A": 8118.13}, "interest_charged": {"Debt A": 9.34}, "payments": {"Debt A": 1367.0}, "ending_balances": {"Debt A": 6760.47}},
          {"month": 16, "target": "Debt A", "opening_balances": {"Debt A": 6760.47}, "interest_charged": {"Debt A": 7.77}, "payments": {"Debt A": 1367.0}, "ending_balances": {"Debt A": 5401.24}},
          {"month": 17, "target": "Debt A", "opening_balances": {"Debt A": 5401.24}, "interest_charged": {"Debt A": 6.21}, "payments": {"Debt A": 1367.0}, "ending_balances": {"Debt A": 4040.45}},
          {"month": 18, "target": "Debt A", "opening_balances": {"Debt A": 4040.45}, "interest_charged": {"Debt A": 4.65}, "payments": {"Debt A": 1367.0}, "ending_balances": {"Debt A": 2678.1}},
          {"month": 19, "target": "Debt A", "opening_balances": {"Debt A": 2678.1}, "interest_charged": {"Debt A": 3.08}, "payments": {"Debt A": 1367.0}, "ending_balances": {"Debt A": 1314.18}},
          {"month": 20, "target": "Debt A", "opening_balances": {"Debt A": 1314.18}, "interest_charged": {"Debt A": 1.51}, "payments": {"Debt A": 1315.69}, "ending_balances": {"Debt A": 0.0}}
        ]
      }
    },
    "recommended_strategy": "tie",
    "total_interest_avalanche": 326.2,
    "total_interest_snowball": 326.2,
    "months_to_debt_free": {"avalanche": 20, "snowball": 20},
    "assumptions": [
      "APR is fixed and divided by 12 for a monthly periodic rate.",
      "Interest is charged before payment and rounded half-up to the nearest cent each month.",
      "Each active debt receives its fixed minimum payment; all remaining budget goes to the strategy target.",
      "Unused budget rolls to the next target in the same month; there are no new charges, fees, or missed payments."
    ],
    "disclaimer": "Educational illustration only; this is not regulated financial advice.",
    "recommendation": "Both methods tie; pay Debt A first and keep the fixed monthly budget."
  }
}

Composition and per-shard licensing

The composition below describes the publishable tagged training file recorded by artifacts/pf-debt-provenance.json.

Shard	Rows	Role	Source	License
pf_debt_payoff	4,488	Recomputable avalanche-versus-snowball payoff comparisons	Original programmatic amortization	CC0-1.0

Published training rows: 4,488. Compilation license: CC0-1.0.

Every published row retains shard, source, and license. The separate holdout is evaluation-only and is not part of the training composition.

Curation methodology

1. Original deterministic scenarios. A fixed random seed generates one to five named debts, integer-cent balances, APR basis points, minimum payments, and a monthly budget that covers every minimum plus an extra amount.
2. Exact amortization labels. Both strategies use the same integer-cent engine. Monthly interest is APR divided by 12 and rounded half-up to the nearest cent. Interest is charged before payment; minimums are paid first; remaining budget targets the highest APR for avalanche or the smallest current balance for snowball; unused budget rolls to the next active debt within the same month.
3. Recomputable outputs. Each completion includes both timelines, total interest, payoff duration, assumptions, a strategy recommendation derived from the computed interest comparison, and an educational-use disclaimer. Long timelines retain the first three months, every twelfth month, and the final three months; shorter timelines retain all months.
4. Deduplication and split isolation. pipeline/dedup_split.py removes exact and near duplicate prompts with normalized hashing and 64-bit token-trigram SimHash before re-splitting. pipeline/provenance.py rejects exact or near train/holdout overlap at maximum Hamming distance 3. The provenance report checked 499 holdout rows and recorded 0 train/holdout leaks.
5. Adaptive Data. Adaptive Data by Adaption Labs scored both the input and adapted data at 9.0, grade A. The unchanged score is reported rather than presented as a gain.

Licensing and attribution

The scenarios, arithmetic labels, and packaging are original work released under CC0-1.0. There is no upstream dataset to attribute. pipeline/provenance.py still attaches source: Original (this work) and license: cc0-1.0 to every row and computes the compilation license as the most restrictive component. With one CC0 component, the compilation is CC0-1.0.

Adaptive Data by Adaption Labs was used for data adaptation.

Intended use

Use CreditCardAPRPath for supervised adaptation, arithmetic-faithfulness evaluation, educational payoff comparisons, and controlled raw-versus-adapted experiments. It is designed to teach models to expose assumptions and produce a trace that can be recomputed independently.

It is not individualized financial advice and must not be used as the sole basis for credit, refinancing, bankruptcy, or other high-stakes decisions.

Limitations

• Scenarios assume fixed APRs, no fees, no new charges, no missed payments, and a fixed monthly budget.
• The engine models monthly periodic interest, not every issuer’s daily-balance method or statement timing.
• Generated balances and budgets do not establish demographic or behavioral representativeness.
• Single-debt cases make avalanche and snowball identical, as the real first-row example shows.
• Exact and SimHash checks do not prove the absence of every semantic near duplicate.

Honest weakness: cent-exact labels are only exact under the documented simplified assumptions. Real accounts can accrue interest daily and include fees, promotional rates, changing minimums, and payment-allocation rules that this dataset does not model.

Reproducibility

Run pipeline/build_pf_dataset.py, then pipeline/dedup_split.py data/pf-debt, then pipeline/provenance.py configs/provenance-pf-debt.yaml. The fixed seed, integer-cent engine, output formatting, provenance tags, license calculation, and leak report are included in this repository.

Created for the Adaption AutoScientist Challenge using Adaptive Data by Adaption Labs.