Morbi is a domain-specialized language model fine-tuned for estate planning, inheritance law, and fiduciary advisory across 47 professional roles.

Model Details

Property	Value
Base Model	Mistral-7B-Instruct-v0.3
Method	QLoRA (4-bit NF4 + LoRA r=64)
Training Data	19,440 examples
Validation Data	823 examples
Training Loss	0.3137
Token Accuracy	95.14%
Training Time	3.3 hours
GPU	NVIDIA RTX A5000 (24GB)
Trainable Params	167M / 3.9B (4.27%)

Training Data Coverage

Accounting & Finance Exams

CFA Level 1 (10 topics), Level 2, Level 3
CPA REG, ISC, AUD, BEC, FAR
CFP Comprehensive
Enrolled Agent (EA)
Series 65, Series 66

Legal Domains

Bar Exam (MEE, MBE, MPT)
Trusts & Estates
Constitutional Law
Real Property
Tax Law
Elder Law
Guardianship
Digital Assets

Usage

python

from peft import PeftModel
from transformers import AutoModelForCausalLM, AutoTokenizer

base = AutoModelForCausalLM.from_pretrained("mistralai/Mistral-7B-Instruct-v0.3")
model = PeftModel.from_pretrained(base, "MorbidCorp/morbi-v1.0.0")
tokenizer = AutoTokenizer.from_pretrained("MorbidCorp/morbi-v1.0.0")

47 Advisor Roles

Estate Attorney, Tax Advisor, Trust Officer, Financial Planner, CPA, Enrolled Agent, Fiduciary, Guardian ad Litem, Probate Judge, Mediator, Elder Law Attorney, Digital Asset Specialist, Insurance Advisor, Real Estate Appraiser, Business Valuator, Forensic Accountant, Charitable Planning Specialist, International Tax Advisor, and more.

License

Apache 2.0