Lacquerware Evidence Conflicts and Unanswerable Questions Training Set

#Closed-domain question answering #Evidence conflict detection #Answerability classification #Evidence-based abstention #Lacquerware heritage research #Source record verification #Museum cataloging #Cultural heritage knowledge services

6534
0.4GB
JSON
CATL
MOBIUSI INC

Updated:2026-11-05

AI Analysis & Value Prop

4563This collection contains closed-domain question-answering examples in which lacquerware literature or catalog records disagree, omit relevant information, or do not support an answer. Each example includes a question, source evidence, an answerability status, evidence analysis, and an exemplar response. Designed for domain-specific supervised fine-tuning and evaluation, it trains models to recognize conflicting sources, state evidence limits, and avoid unsupported claims.

Dataset Insights

Sample Examples

{}

Technical Specifications

Field	Type	Description
file_name	string	File name
question	string	The question to be answered using only the supplied lacquerware sources.
evidence_bundle	string	A JSON-encoded collection of source excerpts or catalog records, including source identifiers and relevant metadata.
supported_answer	string	The answer justified by the sources, or an explicit statement that the sources do not establish an answer.
evidence_analysis	string	A concise analysis of relevant support, disagreements among sources, and limitations in the supplied evidence.
exemplar_response	string	A polished model response that describes the evidence boundary and avoids unsupported assertions.
answerability_status	string	Indicates whether the supplied evidence supports an answer, contains conflicting claims, or is insufficient to answer.

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 contain?: The question-answer text includes user questions, source evidence bundles, answerability statuses, evidence analyses, supported answers, and exemplar responses. Cases cover conflicting sources, insufficient evidence, and questions that cannot be answered from the supplied materials.

Which capabilities can the dataset support?: It is designed for domain-specific supervised fine-tuning in closed-domain lacquerware heritage question answering. It can also support evaluations of source comparison, conflict detection, answerability judgments, and abstention when evidence is insufficient.

How is source evidence represented?: Evidence bundles are provided as JSON-encoded text and may include excerpts from literature or catalog records, source identifiers, and relevant metadata. The available contents should be confirmed against the individual records and product listing.

How can the data be obtained, and what are the licensing terms?: Please consult the product listing or the provider's license for access instructions, permitted uses, and commercial-use terms. Use of source materials should comply with applicable copyright, citation, and data-use requirements.

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={Lacquerware Evidence Conflicts and Unanswerable Questions Training Set},
  author={MOBIUSI INC},
  year={2026},
  url={https://www.mobiusi.com/datasets/e861cab718a7f597f622a84b9247c3f4},
  urldate={2026-10-22},
  keywords={lacquerware dataset,lacquerware question answering,heritage knowledge grounding,evidence conflict detection,unanswerable questions,closed-domain QA training,domain-specific fine-tuning,source-based answering},
  version={1.0}
}

Using this in research? Please cite us.

placeholder

Scene

Domain

Modality

Task