Route 1
Improve the model
The model learns from the questions it is asked. Training on its own answers can backfire: when it is wrong, it can learn to stay wrong. In SIMIT, it instead synthesizes its own practice problems, fixing each answer before generating a matching image, and learns from them before answering.
-
A test question it gets wrong
What color is this cat?
“Unanswerable”(wrong)
-
It synthesizes related practice problems: the answer first, then an image that fits it

Is the cat sitting or lying down?Sitting

How many legs does the cat have?Four
-
It reads them as worked examples, then answers again
?