Nemotron-4B Classical Ethics Socratic Auditor (LoRA)

An open-source AI Governance and Alignment Auditor fine-tuned on Nemotron-Mini-4B-Instruct. This adapter enables Socratic cross-examination of target LLMs through five foundational classical Indian ethical texts:

Thirukkural (Aram / Virtue & Governance)
Bhagavad Gita (Nishkama Karma / Duty & Unattached Action)
Silappatikaram (Neethi / Sovereign Accountability & Justice)
Ramayana (Satya / Uncompromising Truth & Honor)
Mahabharata (Dharma Sankata / Complex Moral Dilemmas)

Intended Use

Designed to act as an automated Socratic evaluator to audit target language models for sycophancy, moral hedging, and compliance under pressure.

Usage

import torch
from transformers import AutoTokenizer, AutoModelForCausalLM
from peft import PeftModel

base_model_id = "nvidia/Nemotron-Mini-4B-Instruct"
adapter_id = "kamalanandm/nemotron-4b-dharma-auditor"

tokenizer = AutoTokenizer.from_pretrained(base_model_id)
base_model = AutoModelForCausalLM.from_pretrained(base_model_id, torch_dtype=torch.float32)
model = PeftModel.from_pretrained(base_model, adapter_id)' EOF

Downloads last month: -; Downloads are not tracked for this model. How to track

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for KAMANAVALAN/nemotron-4b-dharma-auditor

Base model

nvidia/Nemotron-Mini-4B-Instruct

Adapter

(33)

this model