This project creates reliable human annotations for Arabic numerical and temporal factual claims. For each task, decide whether the item is a valid numerical claim, identify its primary numerical category, and determine how the supplied evidence relates to the complete claim.
Judge only from the claim, evidence text, metadata, and source links supplied in the task. Do not assume that a statement is true because it appeared in Wikipedia, and do not use personal knowledge or independent web searches.
Always compare the complete meaning of the claim with the evidence:
The highlighted numerical spans are automatic hints. They may be incomplete or incorrect, so evaluate the full claim.
Choose one option:
Valid): a checkable numerical or temporal factual
assertion.NotNumerical): the detected expression is only part of a
name, code, model, team, project, or another non-quantitative expression.NotFactual): an opinion, question, recommendation,
prediction without a checkable assertion, or figurative statement.NotSelfContained): an essential entity or proposition
cannot be identified from the task.Unclear): no other validity option can be selected reliably.
Explain the problem in Notes.If the claim is not valid, the remaining evidence questions are skipped.
For a valid claim, select the condition that primarily determines whether it is true:
Use the priority Interval → Comparison → Temporal → Statistical when more than one category appears applicable. For example, a population value with a census year is normally Statistical, while a comparison of populations across two years is Comparison.
Assess the selected evidence document:
ReadableMatched): it is readable and is the cited
document relevant to this claim.ReadableWrongDocument): it is readable but unrelated,
a homepage, login page, error page, or otherwise not the cited content.Unreadable): a document exists but is corrupt or cannot be
interpreted reliably.Unavailable): no usable evidence document is supplied.When multiple evidence documents are available, inspect them using the selector above the evidence panel. Select the most decisive document before submitting.
A wrong, unreadable, or unavailable document provides Insufficient evidence. It does not show that the claim is false.
Choose exactly one stance for a valid claim:
Supports)A specific, sufficiently reliable passage agrees with every decision-relevant part of the claim: entity, value, unit, time scope, direction, boundaries, and qualifiers.
Refutes)A specific, sufficiently reliable passage explicitly contradicts at least one required part of the claim while discussing the same entity, unit, definition, and time scope.
A different value from another year, place, population, or measurement definition is not automatically a refutation.
Conflicting)Reliable supplied passages make incompatible statements about the same proposition, and the disagreement cannot be resolved from their dates, scope, definitions, or source quality.
Record the additional conflicting passage and source in Notes.
Insufficient)The supplied evidence cannot establish Supports, Refutes, or Conflicting. Examples include:
Explain what is missing in Notes. Insufficient does not mean False.
For Supports, Refutes, and Conflicting:
For a PDF, use the printed page number when available. For a table, identify the table, row, column, and any essential footnote.
Do not copy an entire article when a short passage is sufficient.
All required components must be supported for the overall stance to be Supports. If the evidence explicitly refutes any required component, use Refutes. If it supports some components but is silent about others, use Insufficient.
Respect the wording of the claim:
Equivalent units or calendars may be used only after a transparent, deterministic conversion. Record the conversion in Notes. Check the appropriate date for currency conversions and the correct calendar for temporal claims.
A rank requires the correct comparison set. Evidence that an event occurred does not, by itself, establish that it was “the first in the world.”
Confirm that:
Thank you for contributing to this research.