# Base build recipe

This describes the base release. Incremental additions are appended without resampling it; current counts and per-source transformations are recorded in the release audit and addition manifests.



- **Canonical recasts.** Each Tasksource task is converted deterministically.
  - Criteria are the source's own label names and answer options.
  - Multiple-choice rows keep every option in a per-row order.
  - A final "all/none of the above" reads "all/none of the other options".
  - Options that cite other options by letter or number keep their order.
  - The question is the task's own when its inputs alone do not say what to predict
    ("What stance does the tweet take on feminism?"), and a generic instruction otherwise.
    Label-verification and packed questions carry it too.
- **Variants.** Low-frequency, deterministic variants cover label verification as `noul`, criterion order, and instruction wording.
- **Packing.** Up to 10% of each classification task's examples are packed, two to four at a time, into `packed_derived` states. Their questions (an item's label, agreement, existence, counts) follow exactly from the gold labels.
- **Mixing (`full`).** Formats get fixed shares of the train rows (47% classification, 30% multiple choice, 3% token labeling, 10% graded (soft-label sources), 10% procedural). Within a format, dataset families get equal shares, scaled by [hand-set weights](https://github.com/sileod/tasksource/blob/main/src/tasksource/metadata/weights.py) (more for adversarial NLI, long documents and preference pairs; less for templated probes), times [audit weights](https://github.com/sileod/tasksource/blob/main/src/tasksource/metadata/audit_weights.py) from a per-task check of Jev on 200 examples: ×1.5 for hard tasks whose gold is right by construction (synthetic logic, theory of mind, spatial reasoning), ×0.5 for near-solved tasks and for hard tasks whose gold is a judgment call (ratings, preferences, crowd sentiment). Sources with many options get slightly more room. Related questions are kept together.
- **Mixing (`default`).** About 1M rows drawn from `full`, never repeating a row. Buckets of related sources get set shares ([jev_mixes.py](https://github.com/sileod/tasksource/blob/main/src/tasksource/metadata/jev_mixes.py): 15% logic, 10% NLI, 10% knowledge QA, 9% long documents and fact-checking, 8% intent and routing, 12% procedural, ...). Inside a bucket, sources get rows by score × √size. The [score](https://github.com/sileod/tasksource/blob/main/src/tasksource/metadata/jev_source_scores.csv) multiplies:
  - correctness: sources that failed label review get 0;
  - the zone of proximal development: how much probability the decision models (Jev, Liquid D1) give the gold answer. Near-solved sources and sources the models miss outright both get less;
  - interest and cleanliness, from Jev yes/no checks for transferable skills, trivial examples and malformed rows;
  - ×0.6 for two-option tasks (about a third of the rows);
  - ×2 for sources picked by reading them.
- **Order and coverage.**
  - The first 1,000 train rows are interleaved to show variety in the Dataset Viewer; the rest is shuffled. Questions of a group stay adjacent throughout.
  - Evaluation benchmarks (BIG-bench, MMLU, BLiMP, MATH test, ...) are left out so they stay clean for evaluation.
- **Sources.** [sources.yaml](sources.yaml) lists every source with its rows, the Hub dataset and revision it was loaded from, the original dataset behind each tasksource copy, and its licenses.
- **Audit trail.** The [source mix](release-audit.json), [failed source list](failed-tasks.json), and [build manifest](build-manifest.json) ship with the data.
- **Reproducible.** The [build runbook](https://github.com/sileod/tasksource/blob/main/docs/jev/README.md) rebuilds the release from [Tasksource](https://github.com/sileod/tasksource)'s [task catalog](https://github.com/sileod/tasksource/blob/main/catalog_english.md).