Indian Flag
Government Of India
A-
A
A+
BhashaBench-Ayur

BhashaBench-Ayur

BhashaBench-Ayur (BBA): Pioneering India’s Ayurvedic AI Benchmark

About Dataset

BhashaBench-Ayur (BBA) is India's first comprehensive benchmark designed to evaluate AI models on traditional Ayurvedic knowledge and practice. As the world's oldest holistic healing system, Ayurveda requires deep understanding of ancient texts, therapeutic principles, herbal medicine, and clinical applications. BBA rigorously tests AI models' ability to comprehend and apply Ayurvedic concepts, drawing from authentic government examinations, institutional assessments, and standardized Ayurvedic education curricula across India. Key Features * Languages: English and Hindi (with plans for more Indic languages including Sanskrit) * Exams: 50+ authentic Ayurvedic government and institutional exams across India * Domains: 15+ specialized Ayurvedic disciplines spanning classical texts to modern practice * Questions: 14,963 rigorously validated, exam-based questions * Difficulty Levels: Easy (7,944), Medium (6,314), Hard (705) * Question Types: Multiple Choice Questions (MCQ), Fill in the Blanks, Match the Column, Assertion-Reasoning * Focus: Traditional knowledge systems, clinical applications, herbal pharmacology, and holistic healthcare approaches

Purpose of Dataset

Bhashabench-ayur Addresses A Critical Gap In Ai Evaluation For Traditional Knowledge Systems. As Interest In Integrative Medicine Grows Globally And India Strengthens Its Traditional Healthcare Sector Through Ayush Initiatives, This Benchmark Provides Essential Tools For: * Evaluating Ai Systems For Traditional Medicine Applications * Developing Culturally-aware Healthcare Ai Solutions * Preserving And Digitizing Ancient Medical Knowledge * Supporting Evidence-based Integration Of Traditional And Modern Medicine

Activity Overview Activity Overview

  • Downloads0
  • Redirect 21
  • Views 94
  • File Size 0

Tags Tags

  • language:hi
  • library:pandas
  • language:en
  • modality:text
  • library:datasets
  • region:us
  • library:polars
  • library:mlcroissant
  • format:parquet
  • license:cc-by-4.0
  • size_categories:10K<n<100K
  • source_datasets:original
  • task_categories:multiple-choice
  • task_categories:question-answering
  • arxiv:2510.25409

License Control License Control

Attribution 4.0 International (CC BY- 4.0)