Yingyaeliae/Hypnos-i1-8B-heretic

Hypnos-i1-8B-heretic: Decensored Quantum-Informed Reasoning Model

This model is a decensored variant of the original adamm-hf/Hypnos-i1-8B, created using the Heretic v1.4.0 tool. It is an 8 billion parameter model with an 8192 token context length, built upon the Nous Hermes 3 (Llama 3.1 8B) architecture.

Key Capabilities

• Specialized Reasoning: Designed for complex logic, chain-of-thought (CoT) reasoning, and mathematical problem-solving, outperforming standard 8B models in these areas.
• Quantum-Informed Training: Uniquely fine-tuned on a dataset enriched with real entropy data generated by IBM Quantum Heron processors (133/156-qubit architecture). This "Quantum Noise Injection" acts as a stochastic regularizer to improve creativity and break deterministic patterns.
• Decensored & Compliant: While based on a robust instruction-following model, this "heretic" version is specifically modified to reduce refusals, with a reported 3/100 refusals compared to the original's 15/100.
• Deep Thinking: Optimized for long-context reasoning, it tends to "think out loud" to ensure higher accuracy on complex queries.

Good For

• Complex Problem Solving: Excels in multi-step logic puzzles and causal inference tasks.
• Creative Generation: The quantum noise injection aims to reduce mode collapse and introduce a unique "temperature" in creative writing.
• Experimental Use Cases: Ideal for researchers and developers exploring the impact of quantum-informed training on LLM behavior.
• Edge & Rapid Prototyping: Its 8B parameter size makes it suitable for deployment on consumer hardware and for quick experimentation.