aifeifei798/Gemma-4-Queen-31B-it
The aifeifei798/Gemma-4-Queen-31B-it is a 31 billion parameter instruction-tuned Gemma-4 model, specifically engineered for superior logical reasoning and physical world modeling. It excels in complex problem-solving scenarios, demonstrating high instruction rigidity and persona stability, outperforming larger models in critical thinking stress tests. This model is particularly suited for applications requiring deep deductive analysis, robust persona role-playing, and precise adherence to complex instructions.
Loading preview...
Model Overview: Gemma-4-Queen-31B-it
The Gemma-4-Queen-31B-it is a 31 billion parameter instruction-tuned model developed by aifeifei798, designed to excel in complex logical reasoning and physical world modeling. Unlike models that prioritize parameter count, Queen-31B focuses on "Logic Density," demonstrating superior performance in stress tests against much larger models (100B+).
Key Capabilities & Differentiators
- Superior Physical World Modeling: Demonstrated 100% accuracy in the "Titan Lab Paradox" test, identifying a hidden physical entry point (heating pipes) and linking it to a timestamped event, where larger models often hallucinate or ignore physical constraints.
- High Instruction Rigidity & Persona Stability: Achieved 75%+ accuracy in the "Steward Protocol" test, maintaining a complex persona and strict adherence to directives (e.g., "no advice") without instruction drift, a common failure point for other LLMs.
- "Purple Spear" Logic Profile: Exhibits a high Logic-to-Token Ratio (LTR), indicating its ability to maintain perfect causal chains under extreme noise and multi-layered constraints, with 100% Spatial Pathing Accuracy and 95% Causal Consistency.
- Specialized Role-Playing: The model is particularly suitable for diverse role-playing scenarios, including "Dark Academia Queen," and offers a "High-Intelligence Persona Matrix" with 50 distinct personas for varied applications.
Ideal Use Cases
- Complex Problem Solving: Excellent for scenarios requiring deep deductive reasoning, such as forensic analysis, engineering-grade problem-solving, and unraveling "locked-room" mysteries.
- Robust Role-Playing & Persona Generation: Highly effective for applications demanding consistent and rigid persona maintenance, from creative writing to interactive simulations.
- Instruction-Following Applications: Suited for tasks where precise adherence to multi-layered instructions and constraints is critical, minimizing "instruction drift."
- Content Generation: Capable of generating detailed writing, songs, and scholarly responses, especially within specific persona contexts like the "Dark Academia Queen."
This model is positioned as a "Forensic Engineer" in its reasoning, reconstructing physical realities rather than generating plausible narratives, making it a significant milestone in AI reasoning for its size.