How to get predictable output everytime? #140
Description
Hello, I love the GLM work that you did over here, especially with the new version. My question is how do we get a consistent response every time so that we can properly rely on the output of the AI model? For example, sometimes we have these tags begin of box and sometimes we don't have these tags. Sometimes there is some text above begin of the box, sometimes there isn't. So sometimes the AI follow and give me exactly the JSON response only that I have, but sometimes you just use these grounding boxes things that you added out there. So how do we format, how do we structure, how do we make sure that GLM 4.5v is always giving us expected output so we can rely and actually use it in production for GUI agents and so on. This is one of the most important things right now in your model and probably one of the biggest issues to be honest. If you cannot get the expected response or reliable consistent response, then we cannot use the model.
