|
Download tasks/task_0010/task.md from Entendre/Crypto-Accounting-Bench: direct link, hf CLI and curl.
- Browser
- Download file 1.65 kB
-
https://huggingface.co/datasets/Entendre/Crypto-Accounting-Bench/resolve/main/tasks/task_0010/task.md
- Command line
-
hf download hf://datasets/Entendre/Crypto-Accounting-Bench/tasks/task_0010/task.md
-
curl -L -o task.md https://huggingface.co/datasets/Entendre/Crypto-Accounting-Bench/resolve/main/tasks/task_0010/task.md
1.65 kB
| # task_0010 | |
| **Benchmark:** crypto-accounting-bench | |
| **Task type:** complete journal entry from transaction evidence | |
| **Organization:** Halden Labs (synthetic) | |
| **Chart of accounts:** 569 accounts | |
| ## What you must do | |
| One crypto transaction is described in `input.json` (and rendered for a model in | |
| `prompt.md`), together with the complete chart of accounts of the organization | |
| that recorded it. Infer the **complete journal entry that organization actually | |
| recorded** - every line, its ledger account, its Debit/Credit side and its | |
| base-currency amount - plus the transaction's asset quantity. | |
| This is a reproduction task, not a best-practice task: the answer is the entry in | |
| that organization's ledger, not the entry you would consider ideal. | |
| ## What you are given | |
| | File | Contents | | |
| |---|---| | |
| | `input.json` | structured public evidence + the organization's chart of accounts | | |
| | `prompt.md` | the same evidence rendered as a ready-to-send model prompt | | |
| | `expected_answer.json` | the graded answer (do not show this to the model) | | |
| ## Answer format | |
| ```json | |
| { | |
| "journalEntry": { | |
| "lines": [ | |
| { | |
| "ledgerAccountName": "<exact ledgerAccountName from the supplied chart>", | |
| "drCr": "Debit | Credit", | |
| "amountBase": "<decimal string, exactly 2 decimal places>", | |
| "currency": "CHF" | |
| } | |
| ] | |
| }, | |
| "assetQuantity": "<the transaction's asset quantity, full precision, decimal string>" | |
| } | |
| ``` | |
| ## Grading | |
| This task is scored against the weighted binary criteria in | |
| `expected_answer.json` -> `rubric`. A task passes at a total weight of | |
| 85% or above. See `../../README.md` for the full evaluation protocol. | |