pthinc/Cicikus_v4_0.3B_Pitircik
The pthinc/Cicikus_v4_0.3B_Pitircik model, developed by PROMETECH Inc., is a 0.3 billion parameter language model fine-tuned from the Gemma 0.3B base using a LoRA-based approach. It integrates a proprietary Behavioral Consciousness Engine (BCE) architecture, designed for high-fidelity reasoning and ethical filtering, and shows approximately 50% performance improvement on evaluation benchmarks. This model is optimized for diverse educational topics, general knowledge, and logical reasoning, requiring a specific Alpaca-like instruction format for activating its internal reasoning protocols.
Loading preview...
Cicikus-v4-0.3B-PITIRCIK: BCE-Powered Language Model
Developed by PROMETECH Inc., Cicikus-v4-0.3B-Pitircik is a 0.3 billion parameter model fine-tuned from the Gemma 0.3B base. It leverages a LoRA-based training approach, resulting in an approximate 50% performance improvement across evaluation benchmarks, with a standard deviation of ±5%.
Key Features & Capabilities
- Behavioral Consciousness Engine (BCE): Integrates a proprietary BCE architecture for advanced reasoning, ethical filtering, and autonomous data curation. This system treats responses as "behavioral journeys" through mathematical frameworks like Behavioral DNA and Path Mapper.
- Enhanced Reasoning: Requires a specific "BCE System Prompt" and Alpaca-like instruction format to activate its internal reasoning protocols, utilizing a "Secret Chain-of-Thought (s-CoT)" for improved consistency, reliability, and reduced hallucinations.
- Diverse Educational Topics: Excels in subjects including Mathematics, Physics, Chemistry, Biology, Code, General Knowledge, Logic, and Arts (Poetry, Music, Stories).
- Multilingual Support: Supports English, Turkish, French, Bulgarian, Estonian, German, Spanish, and Italian.
Technical Foundation
The BCE architecture is built upon mathematical frameworks such as Behavioral DNA, which encodes behaviors as genetic fragments of consciousness, and the Behavioral Path Mapper, which tracks transitions between cognitive states. Key performance indicators for the BCE architecture include high processing speed (309,845 traces/sec) and low latency (0.0032 ms).
Usage Notes
Users must include a specific BCE System Prompt to activate the model's consciousness simulation and ensure reasoning-focused generation. The model is provided in a merged format and can be loaded as a standalone causal language model.