Breaking News Update Reconciliation Question-Answer Dataset

#question answering #temporal fact verification #evidence-based reasoning #breaking news analysis #news update verification #real-time information reconciliation

5000
0.3GB
JSON
CATL
MOBIUSI INC

Updated:2026-10-19

AI Analysis & Value Prop

This dataset features questions that require comparing multiple excerpts from continuously updated breaking news reports, along with their answers. It distinguishes confirmed facts from unconfirmed or disputed claims and records corrections and updates, with evidence citations and reasoning summaries. It supports training and evaluating models that integrate evidence across texts, track changing information, and produce carefully qualified conclusions.

Dataset Insights

Sample Examples

{}

Technical Specifications

Field	Type	Description
file_name	string	File name
reports	array	The source reports and update excerpts provided as inputs for reconciliation.
question	string	The question to answer by comparing the supplied news report excerpts.
confirmed_facts	array	Facts explicitly established or consistently confirmed by the supplied reports.
reasoning_summary	string	A concise explanation of how the excerpts were compared and how the final supported conclusion was reached.
evidence_citations	array	Traceable excerpts that support, contradict, qualify, or correct the answer and claim assessments.
corrections_and_updates	array	Material changes in wording or factual status across successive reports.
current_supported_answer	string	A concise answer limited to conclusions supported by the supplied reports at the latest available update.
unconfirmed_or_disputed_claims	array	Claims whose status remains uncertain, conflicting, corrected, or otherwise not established by the latest reports.

Compliance Statement

Authorization Type	Proprietary - Commercial AI Training License (No Redistribution)
Commercial Use	Requires exclusive subscription or authorization contract (monthly or per-invocation charging)
Privacy and Anonymization	No PII, no real company names, simulated scenarios follow industry standards
Compliance System	Compliant with China's Data Security Law / EU GDPR / supports enterprise data access logs

Frequently Asked Questions

What does the dataset include?: Each example centers on a question that requires comparing multiple news excerpts. It includes a currently supported answer, confirmed facts, unconfirmed or disputed claims, correction and update records, evidence citations, and a reasoning summary.

How does the dataset handle conflicting or unconfirmed reports?: Examples distinguish confirmed facts from claims that remain uncertain, conflict with other reports, or have been corrected. Answers are limited to conclusions supported by the latest available reports, rather than presenting unverified claims as facts.

Which training and evaluation tasks does it support?: The data supports multi-hop question answering, cross-text evidence integration, claim status assessment, update tracking, and evidence citation generation. It can also be used for supervised fine-tuning and evaluation of news reasoning capabilities.

How can I access the data, and what are the licensing terms?: Refer to the product page for the available download method, data format, and licensing terms. Before use, verify that your intended use complies with the license and any requirements applicable to the source news content and related data.

Can't find the data you need?

Post a request and let data providers reach out to you.

Get this Dataset

Verified for Enterprise Use

Cite this Work

@dataset{Mobiusi2026,
  title={Breaking News Update Reconciliation Question-Answer Dataset},
  author={MOBIUSI INC},
  year={2026},
  url={https://www.mobiusi.com/datasets/552ccbce4c32b1504c8c3371306b54bd?dataset_scene_id=17},
  urldate={2026-10-17},
  keywords={breaking news QA dataset,news multi-hop reasoning,report update reconciliation,news fact verification,cross-text evidence integration,news QA training data,correction detection,reasoning SFT dataset},
  version={1.0}
}

Using this in research? Please cite us.

placeholder

Products

Scene

Domain

Modality

Task