The Sam V series began with the purpose to teach a AI an identity, persona and have reasoning. THE EXPERIEMNT is Finished as of 12 May 2025.

Sam-Reason-v3

Model Name: Sam-Reason-v3 Organization: Smilyai-labs License: Apache 2.0 Base Model: Sam-reason-v2 Fine-Tuned By: Smilyai-labs Fine-Tuning Method: Full fine-tuning (no adapters) Token Limit: 8192 Precision: Full precision (fp32)

---

Model Description

Sam-Reason-v3 is a fine-tuned variant of the Sam-reason-v2 model, trained on a custom-crafted dataset that focuses on reasoning-based completions. Each example in the dataset follows a consistent structure:

<prompt>
[reason] <detailed reasoning> [/reason] <final answer>

The model is optimized to:

• Produce coherent multi-step reasoning
• Respond in a consistent and grounded manner
• Maintain high alignment between reasoning and final answers

---

Training Details

• Training platform: Google Colab (T4 GPU)
• Framework: Hugging Face Transformers
• Optimization Strategy: AdamW optimizer, full fine-tuning
• Precision: fp32 (no quantization or LoRA adapters used)
• Training Duration: 6 hours
• Dataset Size: 100,000 examples

---

Dataset Composition

The dataset includes 100,000 samples generated with structured logical questions and aligned answers, using handcrafted reasoning templates for factual, scientific, and common-sense reasoning scenarios.

Each sample includes:

• A clear natural language prompt
• A detailed reasoning section wrapped in [reason]...[/reason]
• A conclusive and matching final answer

Example:

Why do cats purr?
[reason] Cats purr as a way to soothe themselves, especially when they're in a relaxed or content state. [/reason] Purring is a way for cats to calm themselves down when they feel at ease.

---

Intended Use

This model is designed for:

• AI research in multi-step reasoning
• Open-ended educational or tutoring chatbots
• Safe experimentation in logical text generation

Not suitable for:

• Legal or medical advice
• High-stakes decision-making systems

---

License

Apache 2.0

You are free to use, modify, and distribute this model under the terms of the Apache License, Version 2.0.

---

Citation

If you use Sam-Reason-v3 in your work, please cite:

@model{sam_reason_v3,
  title={Sam-Reason-v3: A Reasoning-Tuned Gemma Model},
  author={Smilyai-labs},
  year={2025},
  howpublished={\url{https://huggingface.co/Smilyai-labs/Sam-reason-v3}}
}

---

Contact

For questions or collaborations, contact: [email protected]