tasksource-jev-typed-decisions

Typed decisions from hundreds of curated Tasksource tasks: classification, multiple choice, ordered ratings, probabilities and visual reasoning. Targets come from source annotations or deterministic transformations of them. Native annotation noise and subjective judgments remain.

Load

from datasets import load_dataset

text = load_dataset("tasksource/tasksource-jev-typed-decisions")
vision = load_dataset("tasksource/tasksource-jev-typed-decisions", "vision")
row = text["train"][0]
print(row["state"], row["question"], row["options"], row["target"])

Config	Contents
default	Steered text training mix, plus reviewed incremental additions; validation and test included.
full	Larger text pool with per-source caps, including incremental additions; validation and test included.
filtered-full	Filtered, training-only historical snapshot. Later additions are not included. See selection policy and counts.
vision	Multimodal decisions with ordered images and text choices. See visual sources, sampling and provenance.

Exact split sizes are in the Dataset Viewer metadata. Task coverage and counts are in sources.yaml and vision/sources.yaml.

Row format

Field	Meaning
state	Text to reason about.
question	The decision to make.
kind	choice, score or noul.
options	Criteria in their presented order; empty for noul.
target	Distribution aligned with options, or [p_true] for noul.
id, group_id, question_id	Decision identity and grouping of related questions.
example_id	Content-derived example identity used by the build.
source_index	Original row of the example in its source dataset, as split:row (the split before any re-splitting); empty for packed and generated rows.
source, split, variant	Task identity, split annotation and recast variant.
license, license_use	Recorded source terms and their usage classification.
images	Vision only: ordered Hugging Face image features, decoded on access.
metadata	Vision only: JSON-string source identity, geometry, augmentation and license evidence.

{"state": "My body cast a shadow over the grass. What was the cause of this?",
 "question": "Choose the criterion that best answers the question.",
 "kind": "choice", "options": ["The sun was rising.", "The grass was cut."],
 "target": [1.0, 0.0], "source": "super_glue/copa"}

choice selects a criterion; score uses an ordered scale. Soft targets preserve annotator disagreement or interpolate numerical ratings between neighboring scale anchors. noul represents the probability that a proposition is true. Scale meaning is determined by the question and criteria.

Rows sharing group_id can be assembled into a multi-question request. Vision images stay separate from text and must be passed through the model's image adapter; answer options remain text. See AGENTS.md for request assembly and row semantics.

Construction and evaluation

Tasksource standardizes native labels and choices, adds task-specific questions where needed, and applies deterministic option permutations and low-frequency text variants. Some text rows pack related decisions over a shared state. A decision's majority label keeps at most three rows per row of its other labels (graded answers count as yes above 0.5; a question with a single label is dropped), so skewed binary sources such as answer selection or toxicity no longer reward always answering the common label; procedural questions keep each label under 60%. The visual config contains direct decisions and shares sampling budgets across related task views.

Native source partitions are retained where available; derived group-based holdouts and overlap exclusions are documented per source. These checks do not establish absence of all cross-dataset or benchmark contamination. For example, Mind2Web text tasks use derived holdouts from public training trajectories, not official benchmark tests; alignment with private Decision Index evaluation rows is unverified. Mind2Web is excluded from the visual config.

Source and recast reviews are coarse checks, not corpus-wide correctness guarantees. English examples illustrate the exported and canonical representations.

Licenses

This collection has mixed source terms, hence license: other. Original annotation and image terms still apply. license_use is a best-effort classification: commercial, non-commercial or unspecified; it is not legal clearance. Missing or scoped evidence is retained explicitly.

text = text.filter(lambda use: use == "commercial", input_columns="license_use")

Visual annotation licenses may not cover third-party images. The visual license review records unresolved image rights and explicit restrictions, including FigureQA's prohibition on redistribution. See component-level evidence and the vision documentation before reuse.

Documentation and provenance

• Text sources and revisions, release audit, build manifest and failed sources.
• Vision documentation and filtered-full selection evidence.
• Build runbook, base build recipe and task catalogs.
• Release history. Addition manifests and sampled reviews retain their original snapshot context.

Citation

@inproceedings{sileo-2024-tasksource,
  title = {tasksource: A Large Collection of {NLP} tasks with a Structured Dataset Preprocessing Framework},
  author = {Sileo, Damien},
  booktitle = {Proceedings of the 2024 Joint International Conference on Computational Linguistics, Language Resources and Evaluation (LREC-COLING 2024)},
  year = {2024},
  pages = {15655--15684},
  url = {https://aclanthology.org/2024.lrec-main.1361/}
}