gnumanth/mahabharata-characters

Mahabharata Multidimensional Characters Dataset

An analytical, multidimensional dataset classifying 307 figures from the Sanskrit epic Mahabharata, combining genealogical and martial attributes with TypeSafe AI System One decision primitives (Choice, Score, Noul) and deterministic ethical indices.

Dataset Summary

• Total Characters: 307
• Factions: 90 Pandava Coalition, 153 Kaurava Host, 64 Neutral / Rishis / Allied Kings
• Philosophical Gunas: 135 Sattva (Purity / Divine Wisdom), 19 Rajas (Passion / Ambition), 153 Tamas (Inertia / Darkness)
• Philosophical Authority: Grounded in Śrī Madhvāchārya's Mahābhārata Tātparya Nirṇaya (MBTN), providing the definitive ontological classification of epic figures based on intrinsic swabhava, celestial origins (Taratamya), and demonic avesha.
• Source Lore Grounding: Śrī Madhvāchārya's Mahābhārata Tātparya Nirṇaya, Critical Edition of the Mahābhārata (Bhandarkar Oriental Research Institute, Pune), and Harivaṃśa.
• Model Evaluation: TypeSafe AI jev-latest System One calibrated decision engine.

Key Classifications Grounded in Śrī Madhvāchārya's MBTN

1. Bhimasena (bhima) — Sattva (Supreme Jivottama):
   • Second incarnation of Mukhya Prana (Vayu), foremost among all Jivas (Jivottama).
   • Characterized by supreme Jnana, Bhakti, Vairagya, and invincible Bala.
   • Every conquest (Jarasandha, Kichaka, Bakasura, Hidimba, Duryodhana, and all 100 Kauravas) performed as a holy Yajna without personal anger, completely free of adharma.
2. Karna (karna) — Tamas (Sahasrakavacha Daitya Avesha):
   • Although born of Surya, Karna was dominated by the demonic avesha of Sahasrakavacha (Dambhodbhava Daitya).
   • Driven by deep Hari-dvesha and Pandava-dvesha, instigator of Draupadi's attempted disrobing and cowardice in the slaughter of unarmed Abhimanyu.
3. Draupadi (draupadi) — Sattva (Bharati Devi Avatara):
   • Incarnation of Bharati Devi (consort of Mukhya Prana), supreme female divinity after Sri Lakshmi.
   • Embodiment of unblemished moral righteousness and spiritual wisdom.
4. Duryodhana & the 100 Kauravas — Tamas (Kali Purusha & Asuras):
   • Duryodhana is the direct human incarnation of Kali Purusha; Duhshasana is Duhsaha; all 100 brothers represent Krodhavasha Asuras/Rakshasas (MBTN Chapter 18).
5. Dronacharya (dronacharya) — Sattva (Brihaspati Avatara):
   • Incarnation of Brihaspati, Guru of Devas; bound to the Kaurava court by royal sustenance (Kuru-anna) while his heart remained devoted to Dharma and Pandavas.
6. Shalya (shalya) — Tamas (Sanhlada Daitya Avatara):
   • Incarnation of Sanhlada Daitya, demonic brother of Prahlada who fought on the side of evil.
7. Ekalavya (ekalavya) — Tamas (Krodhavasha Asura Amsha):
   • Incarnation of a Krodhavasha Asura, fierce ally of Jarasandha against Krishna and Dwaraka.
8. Hanuman (hero-hanuman-mahabharata) — Sattva (Mukhya Prana Prathama Avatara):
   • First incarnation of Mukhya Prana, chiranjivi hero, elder divine brother of Bhima, and banner protector of Arjuna's chariot.

Dataset Structure

Data Files

• data/train.parquet: Optimized columnar Parquet split.
• data/train.jsonl: JSON Lines format for JSON-based pipelines.
• data/characters.csv: Tabular CSV format.

Fields

Field Name	Type	Description
id	string	Canonical normalized identifier (e.g. arjuna, karna, ekalavya)
name	string	Transliterated name in English
sanskrit_name	string	Devanagari Sanskrit title (e.g. अर्जुन, एकलव्य)
epithets	list[string]	Traditional titles and warrior designations
role	string	Narrative function, kingship, or military role
gender	string	Gender (Male, Female, Non-binary)
dynasty_kingdom	string	Kingdom of origin (e.g. Kuru, Yadava, Panchala, Matsya, Nishada)
father	string	Father's name
mother	string	Mother's name
allegiance	string	War allegiance (pandava, kaurava, neutral)
primary_weapon	string	Iconic weapon or astra
special_powers	list[string]	Divine boons, celestial astras, and warrior feats
notable_events	list[string]	Key narrative deeds across the epic
death_or_fate	string	Fate at Kurukshetra, Svargarohana, or post-war reign
background	string	Comprehensive narrative biography
moral_dilemma	string	Philosophical conflict (Dharma vs. Sookshma Dharma, Vows vs. Conscience)
dominant_guna	string	TypeSafe Choice judgment: sattva, rajas, or tamas
guna_confidence	float	Calibrated confidence of Guna distribution (0.0–1.0)
guna_prob_sattva	float	Calibrated probability of Sattva
guna_prob_rajas	float	Calibrated probability of Rajas
guna_prob_tamas	float	Calibrated probability of Tamas
dharmic_flexibility_score	float	Continuous Score from 0.0 (Adharma) to 4.0 (Inflexible Legalism)
martial_prowess_score	float	Continuous Score from 0.0 (Non-Combatant) to 4.0 (Maharatha/Titan)
tragic_flaw	string	Primary existential flaw archetype
subverted_dharma_in_war_noul	float	Calibrated probability (0.0–1.0) of violating codes during war
bound_by_tragic_vow_noul	float	Calibrated probability (0.0–1.0) of entrapment by vows/curses
attained_liberation_or_svarga_noul	float	Calibrated probability (0.0–1.0) of spiritual liberation/Svarga
composite_dharmic_integrity	int	Algorithmically weighted dharma index (0–100)
composite_martial_dominance	int	Algorithmically weighted combat dominance index (0–100)
composite_tragic_index	int	Algorithmically weighted fate burden index (0–100)
composite_cosmic_influence	int	Average of dharmic integrity and martial dominance

Quickstart with datasets

from datasets import load_dataset

# Load dataset directly from Hugging Face Hub
ds = load_dataset("gnumanth/mahabharata-characters", split="train")

# Inspect a record
print(ds[0]["name"], "->", ds[0]["dominant_guna"])

# Filter Tamasic figures (e.g. Duryodhana, Shakuni, Ekalavya, Jarasandha)
tamasic = ds.filter(lambda x: x["dominant_guna"] == "tamas")
print(f"Total Tamasic characters: {len(tamasic)}")

Intended Uses

• Comparative Literary & Ethical Analysis: Computational studies of Sanskrit epic character dynamics.
• Decision Model Benchmarking: Evaluating LLM consistency against calibrated System One probabilistic decisions.
• Narrative & RPG Generation: Grounding knowledge graphs and agents in epic character attributes.

Citation

@dataset{mahabharata_classifier_2026,
  author    = {Hemanth HM},
  title     = {Mahabharata Multidimensional Characters Dataset: 300+ Figures with TypeSafe AI System One Dimensions},
  year      = {2026},
  publisher = {Hugging Face},
  howpublished = {\url{https://huggingface.co/datasets/gnumanth/mahabharata-characters}}
}