Multi-Turn Dialogue Dataset for Insufficient Knowledge Evidence Scenarios
- Records
- 4260
- Size
- 0.4GB
- Format
- JSON
- Provider
MOBIUSI INC- Updated
- v1.02026-11-05
Covers conversations in which knowledge-base evidence is missing, does not support a user's conclusion, or supports only a partial answer. Records pair multi-turn dialogue text with related evidence and show how an assistant can state limitations, avoid unsupported claims, and continue through clarification or a clearly scoped response. The dataset supports reliable response training and evaluation for RAG knowledge-base QA and dialogue supervised fine-tuning.
Overview
Dataset Insights
Preview
Sample Examples
Schema
Technical Specifications
| Field | Type | Description |
|---|---|---|
| file_name | string | File name |
| source_dialogue | string | The original multi-turn dialogue, including speaker labels and assistant responses. |
| source_evidence | string | The knowledge-base evidence associated with the dialogue; the value may be empty when no evidence is available. |
| conversation_turns | array | The dialogue parsed into an ordered sequence of speaker turns. |
| evidence_assessment | object | A structured assessment of what the supplied evidence supports and does not support. |
| assistant_strategies | array | Reliable response behaviors demonstrated by the assistant. |
| clarification_needed | boolean | Indicates whether a follow-up question is needed to provide a supported or more complete answer. |
Trust
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
FAQ
Frequently Asked Questions
- What does the dataset contain?
- It includes multi-turn dialogue text and associated knowledge evidence, along with structured information on dialogue turns, evidence support, assistant strategies, and whether clarification is needed.
- Which capabilities can it support?
- It can support training models to detect missing or insufficient evidence, avoid claims beyond the available evidence, communicate uncertainty, and ask clarifying questions when needed. It can also be used to evaluate these capabilities.
- How can I access the data, and what are the specifications?
- Check the data provider's product page for available download or delivery options. The listed fields cover dialogues, evidence, and structured assessments; record counts, file formats, and delivery specifications should be confirmed on the product page or in the delivery details.
- What licensing applies to the dataset?
- Permitted uses, commercial rights, and redistribution restrictions are governed by the license listed on the product page or an agreement with the provider. Confirm applicable terms before using the data for training or evaluation.
Can't find the data you need?
Post a request and let data providers reach out to you.
Get this Dataset
High-quality data, ready for your next AI breakthrough.
Cite this Work
@dataset{Mobiusi2026,
title={Multi-Turn Dialogue Dataset for Insufficient Knowledge Evidence Scenarios},
author={MOBIUSI INC},
year={2026},
url={https://www.mobiusi.com/datasets/8d998ba2c54f3c8505a31abbd0210f1c?dataset_scene_id=19},
urldate={2026-10-26},
keywords={insufficient evidence dialogue,multi-turn dialogue dataset,RAG reliable responses,evidence-grounded generation,knowledge base QA training,dialogue SFT data,uncertainty communication training,clarification question generation},
version={1.0}
}Using this in research? Please cite us.