faunix/LiarAI
LiarAI-2B by faunix is a 2.3 billion parameter causal language model, based on Qwen/Qwen3.5-2B, specifically engineered to produce highly inaccurate and hallucinatory responses. With a context length of 32768 tokens, this model is uniquely designed for generating creative falsehoods and nonsensical outputs. It serves as an experimental model to explore extreme hallucination, making it distinct from conventional LLMs aiming for factual accuracy.
Loading preview...
Overview
LiarAI-2B, developed by faunix, is a 2.3 billion parameter language model built upon the Qwen/Qwen3.5-2B architecture. Unlike typical large language models that strive for factual correctness, LiarAI-2B is intentionally designed to "hallucinate in approximately 101% of cases," making it a unique exploration into controlled inaccuracy. It features a substantial context length of 32768 tokens.
Key Capabilities
- Extreme Hallucination: Engineered to consistently generate false and nonsensical information.
- Creative Misinformation: Provides imaginative and often humorous incorrect answers to a wide range of prompts.
- Experimental Design: Serves as a demonstration of a model optimized for generating untruths, rather than factual data.
Usage Recommendations
For optimal "liar" performance, the model is recommended to be used with specific generation parameters:
- Temperature: 1.2
- Top-P: 0.95
- Top-K: 20
Intended Use Cases
This model is not intended for applications requiring factual accuracy or reliability. Instead, it is ideal for:
- Humorous Content Generation: Creating amusingly incorrect statements or stories.
- April Fools' Day Content: Generating content specifically designed to be misleading or funny.
- Research into Hallucination: Studying the mechanisms and outputs of models intentionally designed to hallucinate.
- Creative Writing Prompts: Inspiring unconventional narratives through deliberately false information.