48 Downloads Updated 2 months ago
This model was fine tuned with GRPO using the ZeroGPT as the reward function, making it extremely effective at evading the ZeroGPT and GPTZero classifiers.
Model performance is not guaranteed on other AI detection tools.
The model on Ollama is quantized to Q8_0.
Run this command in your terminal.
ollama run Beanie/Trojan-1-Preview-Qwen3-1.7B
Some variations from the default system prompt have been tested and do affect how the model performs against the classifier.