manuelcaccone/gemma-3-actuaryEnough2

👋 Connect with me on LinkedIn!

Manuel Caccone - Actuarial Data Scientist & Open Source Educator

Let's discuss actuarial science, AI, and open source projects!

---

🤖 Gemma-3 ActuaryEnough2: Bringing Actuarial AI to Everyone

---

🚩 Model Description

Gemma-3-actuaryEnough2 is a fine-tuned Gemma-3-270M model trained on over 11,000 actuarial question-answer pairs, purpose-built to translate simple insurance queries into rigorous actuarial technical language. It powers ActuaryEnough and is released as open source for educational and research use.

---

✨ Key Features

• 🎯 Domain-specific: Focused exclusively on actuarial and insurance Q&A.
• 📚 Educational: Makes complex actuarial terminology accessible for all users.
• 🚀 Efficient: Fine-tuned with Unsloth for rapid, scalable training.
• 🔓 Open Source: Apache 2.0 License; easy to reuse, adapt, remix.
• 🌐 Widget & Demo: Integrated as a live demo on ActuaryEnough.

---

💡 Intended Use Cases

• Education: For students and actuaries in training, or for professionals retraining in actuarial language.
• Translation: Make practical insurance questions understandable at professional actuarial level.
• Research: Support for actuarial research, Q&A, and domain adaptation.

Examples

# Premium Calculation Example
Input: "How much should I pay for car insurance? Rephrase:"
Output: "This relates to premium calculation considering risk factors such as exposure units, loss frequency, severity distributions, and loading factors for expenses and profit margins."

---

📂 Training Data

• Primary Dataset: actuarial-qa-11k - Over 11,000 manually curated actuarial question–answer pairs
• Specialized Dataset: actuary-enough-qa-dataset - Actuarial question simplification examples
• Topics: Life and non-life insurance, risk assessment, regulation, reserves, actuarial mathematics, terminology simplification
• Language: English
• Format: Instruction-following format optimized for text generation tasks

---

📊 Training Statistics

Metric	Value / Range	Notes
Epochs	~51	Reached at end of training
Global Steps	>68,000
Initial Train Loss	~2.2	At start
Final Train Loss	~1.4	At end
Learning Rate	8e-7 → ≈0	Linear decay throughout training
Gradient Norm	5 – 15	Generally stable with rare spikes
Hardware	RTX 3090, 16-core CPU	24GB VRAM, 94GB RAM, CUDA 12.8, Linux 6.1

---

🛠️ Dependencies

Python 3.12.11
transformers
torch
unsloth
wandb==0.21.1
pydantic==2.11.7
# ...for full list, check requirements.txt

---

⚠️ Limitations & Ethics

• No pricing or decision support: For education and inspiration only, not for real insurance contracts.
• Not a substitute for an actuary: Always consult professionals for real-world decisions.
• Coverage: Designed and tested specifically for the insurance/actuarial domain.
• Training data bias: Outputs may reflect source content.

---

💻 Usage Example

from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("manuelcaccone/gemma-3-actuaryEnough2")
model = AutoModelForCausalLM.from_pretrained("manuelcaccone/gemma-3-actuaryEnough2")

prompt = "Which factors determine life insurance premiums?"
toks = tokenizer(prompt, return_tensors="pt")
output = model.generate(**toks, max_new_tokens=256, temperature=0.7)
print(tokenizer.decode(output[0], skip_special_tokens=True))

---

🌟 Related Datasets

This model is part of the ActuaryEnough ecosystem and uses multiple specialized datasets:

• Primary Training: actuarial-qa-11k - Complete dataset with 11,000+ actuarial Q&A pairs
• Specialized Component: actuary-enough-qa-dataset - Question simplification dataset
• Demo App: ActuaryEnough - Interactive web application
• Model Repository: gemma-3-actuaryEnough2 - This fine-tuned model

---

👤 Author & Citation

• Creator: Manuel Caccone (Actuarial Data Scientist & Open Source Educator)
• LinkedIn · manuel.caccone@gmail.com

@model{caccone2025actuaryenough,
  title={Gemma-3 ActuaryEnough2: A Fine-tuned Model for Actuarial Education},
  author={Caccone, Manuel},
  year={2025},
  publisher={Hugging Face},
  url={https://huggingface.co/manuelcaccone/gemma-3-actuaryEnough2},
  note={Educational model for actuarial science and insurance terminology}
}

---

📜 License

Apache 2.0 License — use, modify, and cite for ethical, research, and educational purposes.

---

🤝 Want to collaborate or discuss actuarial AI?

---

Part of the ActuaryEnough open-source education initiative—bringing actuarial science closer to everyone!