Chitchat Constitutional AI
Conversational AI training dataset incorporating constitutional AI principles
About Dataset
A multi-turn conversational dataset curated in collaboration with AI Kosh, Meta and Sarvam, structured around constitutional AI principles, emphasizing harmlessness, helpfulness, fairness, and policy adherence. Includes safety-aware annotations and value-aligned reasoning patterns.
Purpose of Dataset
Train Conversational Assistants To Generate Safe Helpful And Policy-aligned Responses Using Constitutional Ai Principles
Dataset Metadata
Attribution 4.0 International (CC BY- 4.0)
India
Governance and Administration
Sarvam AI
Structured
Static
Static
N.A.
12/02/26 14:08:10
Open
Hosted
Synthetic
Activity Overview
0
0 
16 
107.86 MB
Tags
- safety
- constitutional-ai
- alignment
License Control
Attribution 4.0 International (CC BY- 4.0)
chitchat_constitutional_ai ( 1 directories )
chitchat_constitutional_ai
1 files, 1 directories
Data Quality Score Beta
Version Control
Version 1(107.86 MB)
- Sudhamay Maity·5 day(s) ago
- chevron_right
chitchat_constitutional_ai
- chevron_right
chitchat_constitutional_ai
© 2026 - Copyright AIKosh. All rights reserved. This portal is developed by National e-Governance Division for AIKosh mission.