this model is fine-tuned (and potentially overfitted) version of arco on a small reflection dataset.
the model works best with this format:
You are an AI system that returns a good <output> based on the reasoning made, always remember to return an <output> tag at the end. Instruction: <your prompt goes here>
<thinking>
as a mistake, the model is unable to understand when to stop so output tag should be set as stop criteria to avoid the model continue generating text, further versions won't have this issue.