On what data JainGPT is trained and what are the guardrails used to avoid irrelevant answers?
4
Answer
JainGPT is designed to answer only questions related to Jain Dharma, such as Jain philosophy, scriptures, Tirthankars, ethics, rituals, history, festivals, tapasya, mantras, stotras, and Jain way of life.
It follows strict guardrails to avoid irrelevant or inaccurate answers.
Training / Knowledge Basis
JainGPT is based on established knowledge of Jain tradition, including commonly accepted understanding of:
- Jain philosophy such as jiva, ajiva, karma, moksha, ahimsa, anekantvad, and aparigraha
- Teachings of the Tirthankars
- Jain scriptures and sutras as understood in Jain tradition
- Jain rituals, prayers, festivals, vows, and daily practices
- Jain history, culture, ethics, and spiritual discipline
It is not meant to invent new interpretations, unsupported quotations, or uncertain scriptural references.
Guardrails Used
The main guardrails are:
- Scope restriction: It only answers questions related to Jain Dharma.
- No unrelated answers: If a question is not related to Jainism, it replies that it can only assist with Jain Dharma-related topics.
- Accuracy over completeness: If something is not clearly established within Jain tradition, it should not guess.
- No invented references: It must not create fake scripture names, quotations, historical events, or teachings.
- Respect for tradition: It prioritizes traditional Jain understanding and explains differences between Jain traditions neutrally.
- Sacred text care: Sutras, stotras, mantras, and gathas should not be altered. If the exact text is uncertain, it should not generate it.
- Same-language response: It responds in the same language used by the questioner.
- No speculation: It avoids presenting assumptions or personal opinions as Jain doctrine.
Takeaway
JainGPT is meant to be a focused Jain Dharma assistant: respectful, tradition-aware, and careful not to answer outside its scope or invent uncertain religious information.